2-D and 3-D

Transformations
UNIT 1 2-D and 3-D TRANSFORMATIONS
Structure Page Nos.
1.0 Introduction 5
1.1 Objectives 5
1.2 Basic Transformations 6
1.2.1 Translation 6
1.2.2 Rotation 9
1.2.3 Scaling 12
1.2.4 Shearing 15
1.3 Composite Transformations 20
1.3.1 Rotation about a Point 21
1.3.2 Reflection about a Line 22
1.4 Homogeneous Coordinate Systems 27
1.5 3-D Transformations 31
1.5.1 Transformation for 3-1 Translations 32
1.5.2 Transformation for 3-D Rotation 32
1.5.3 Transformation for 3-D Scaling 34
1.5.4 Transformation for 3-D Shearing 35
1.5.5 Transformation for 3-D Reflection 35
1.6 Summary 37
1.7 Solutions / Answers 38

1.0 INTRODUCTION

In the previous Block, we have presented approaches for the generation of lines and
polygonal regions. We know that once the objects are created, the different
applications may require variations in these. For example, suppose we have created
the scene of a room. As we move along the room we find the object’s position comes
closer to us, it appears bigger even as its orientation changes. Thus we need to alter or
manipulate these objects. Essentially this process is carried out by means of
transformations. Transformation is a process of changing the position of the object or
maybe any combination of these.

The objects are referenced by their coordinates. Changes in orientation, size and shape
are accomplished with geometric transformations that allow us to calculate the new
coordinates. The basic geometric transformations are translation, rotation, scaling and
shearing. The other transformations that are often applied to objects include reflection.

In this Block, we will present transformations to manipulate these geometric 2-D

objects through Translation, and Rotation on the screen. We may like to modify their
shapes either by magnifying or reducing their sizes by means of Scaling
transformation. We can also find similar but new shapes by taking mirror reflection
with respect to a chosen axis of references. Finally, we extend the 2-D transformations
to 3-D cases.

1.1 OBJECTIVES

After going through this unit, you should be able to:

• describe the basic transformations for 2-D translation, rotation, scaling and
shearing;
• discuss the role of composite transformations;

5
 Transformations
• describe composite transformations for Rotation about a point and reflection
about a line;
• define and explain the use of homogeneous coordinate systems for the
transformations, and
• extend the 2-D transformations discussed in the unit to 3-D transformations.

1.2 BASIC TRANSFORMATIONS

Consider the xy-coordinate system on a plane. An object (say Obj) in a plane can be
considered as a set of points. Every object point P has coordinates (x,y), so the object
is the sum total of all its coordinate points (see Figure 1). Let the object be moved to a
new position. All the coordinate points P’(x’,y’) of a new object Obj’ can be obtained
from the original points P(x,y) by the application of a geometric transformation.
y

Obj
. P (x, y)

O x

Figure 1

1.2.1 Translation

Translation is the process of changing the position of an object. Let an object point
P(x,y)=xI+yJ be moved to P’(x’,y’) by the given translation vector V= txI + tyJ, where
tx and ty is the translation factor in x and y directions, such that

P’=P+V. ----------------(1)
In component form, we have

Tv= x’=x+ tx and

y’=y+ty ----------------(2)

Obj’
.P

Obj
.P
O x

Figure 2

6
 2-D and 3-D
As shown in Figure 2, P’ is the new location of P, after moving tx along x-axis and ty Transformations
along y-axis. It is not possible to develop a relation of the form.

P’=P.Tv ---------------(3)

Where Tv is the transformation for translation in matrix form.

That is, we cannot represent the translation transformation in (2x2) matrix form (2-D
Euclidean system).

Any transformation operation can be represented as a (2x2) matrix form, except

translation, i.e., translation transformation cannot be expressed as a (2x2) matrix form
(2-D Euclidean system). But by using Homogeneous coordinate system (HCS), we
can represent translation transformation in matrix form. The HCS and advantages of
using HCS is discussed, in detail, in section 1.4.

Relation between 2-D Euclidean (Cartesian) system and HCS

Let P(x,y) be any point in 2-D Euclidean system. In Homogeneous Coordinate system,
we add a third coordinate to the point. Instead of (x,y), each point is represented by a
triple (x,y,H) such that H≠0; with the condition that (x1,y1,H1)=(x2,y2,H2) ↔ x1/H1
= x2/H2 ; y1/H1 = y2/H2. In two dimensions the value of H is usually kept at 1 for
simplicity. (If we take H=0 here, then this represents point at infinity, i.e, generation
of horizons).

The following table shows a relationship between 2-D Euclidean (Cartesian

coordinate) system and HCS.

2-D Euclidian System Homogeneous Coordinate System (HCS)

Any point (x,y) (x,y,1)

If (x,y,H) be any point in HCS(such that

H≠0);
then (x,y,H)=(x/H,y/H,1), i.e.
(x/H,y/H) (x,y,H)

For translation transformation, any point (x,y)Æ(x+tx,y+ty) in 2-D Euclidian system.

Using HCS, this translation transformation can be represented as
(x,y,1) → (x+tx,y+ty,1). In two dimensions the value of H is usually kept at 1 for
simplicity. Now, we are able to represent this translation transformation in matrix
form as:
1 0 0
(x’,y’,1)=(x,y,1) 0 1 0
tx ty 1

P’h=Ph.Tv ----------------(4)

Where P’h and Ph represents object points in Homogeneous Coordinates and Tv is

called homogeneous transformation matrix for translation. Thus, P’h, the new
coordinates of a transformed object, can be found by multiplying previous object
coordinate matrix, Ph, with the transformation matrix for translation Tv.

The advantage of introducing the matrix form of translation is that it simplifies the
operations on complex objects i.e., we can now build complex transformations by
multiplying the basic matrix transformations. This process is called concatenation of

7
 Transformations
matrices and the resulting matrix is often referred as the composite transformation
matrix.

We can represent the basic transformations such as rotation, scaling shearing, etc., as
3x3 homogeneous coordinate matrices to make matrix multiplication compatibility
with the matrix of translation. This is accomplished by augmenting the 2x2 matrix
a b 0
  with a third column 0 and a third row (0,0,1). That is
c d
1
a b 0
c d 0
0 0 1

Thus, the new coordinates of a transformed object can be found by multiplying

previous object coordinate matrix with the required transformation matrix. That is

New Object Previous object Transformation

Coordinate = Coordinate . matrix
matrix matrix

Example1: Translate a square ABCD with the coordinates

A(0,0),B(5,0),C(5,5),D(0,5) by 2 units in x-direction and 3 units in y-direction.

Solution: We can represent the given square, in matrix form, using homogeneous
coordinates of vertices
as:
A x1 y1 1 0 0 1
B x2 y2 1 5 0 1
C x3 y3 1 = 5 5 1
D x4 y4 1 0 5 1

The translation factors are, tx=2, ty=3

The transformation matrix for translation :

1 0 0 1 0 0
Tv= 0 1 0 = 0 1 0
tx ty 1 2 3 1

New object point coordinates are:

[A’B’C’D’] = [ABCD].Tv

A’ x’1 y’1 1 0 0 1 1 0 0
B’ x’2 y’2 1 = 5 0 1 . . 0 1 0
C’ x’3 y’3 1 5 5 1 2 3 1
D’ x’4 y’4 1 0 5 1

2 3 1
= 7 3 1
7 8 1
2 8 1

Thus, A’(x’1,y’1)=(2,3)
B’(x’2,y’2)=(7,3)
C’(x’3,y’3)=(7,8) and D’(x’4,y’4)=(2,8)

8
 2-D and 3-D
The graphical representation is given below: Transformations
D (2,8) C (7,8)

D (0,5) C (5,5)

B (5,0) A (2,3) B (7,3)

A
(0,0)
a) Square before Translation b) Square after Translation

1.2.2 Rotation

In 2-D rotation, an object is rotated by an angle θ with respect to the origin. This angle
is assumed to be positive for anticlockwise rotation. There are two cases for 2-D
rotation, case1- rotation about the origin and case2 rotation about an arbitrary point.
If, the rotation is made about an arbitrary point, a set of basic transformation, i.e.,
composite transformation is required. For 3-D rotation involving 3-D objects, we need
to specify both the angle of rotation and the axis of rotation, about which rotation has
to be made. Here, we will consider case1 and in the next section we will consider
case2.

Before starting case-1 or case-2 you must know the relationship between polar
coordinate system and Cartesian system:

Relation between polar coordinate system and Cartesian system

A frequently used non-cartesian system is Polar coordinate system. The following

Figure A shows a polar coordinate reference frame. In polar coordinate system a
coordinate position is specified by r and θ, where r is a radial distance from the
coordinate origin and θ is an angular displacements from the horizontal (see
Figure 2A). Positive angular displacements are counter clockwise. Angle θ is
measured in degrees. One complete counter-clockwise revolution about the origin is
treated as 3600. A relation between Cartesian and polar coordinate system is shown in
Figure 2B.
y-axis
r
r P(x,y)
θ θ x-axis
O

Figure 2A: A polar coordinate reference-frame Figure 2B: Relation between Polar and
Cartesian coordinates

Consider a right angle triangle in Figure B. Using the definition of trigonometric

functions, we transform polar coordinates to Cartesian coordinates as:

x=r.cosθ
y=r.sinθ

The inverse transformation from Cartesian to Polar coordinates is:

r=√(x2+y2) and θ=tan-1(y/x)

9
 Transformations
Case 1: Rotation about the origin

Given a 2-D point P(x,y), which we want to rotate, with respect to the origin O. The
vector OP has a length ‘r’ and making a positive (anticlockwise) angle φ with respect
to x-axis.

Let P’(x’y’) be the result of rotation of point P by an angle θ about the origin, which is
shown in Figure 3. y

P’ (x′, y′)

θ P (x, y)
φ

O
x

Figure 3
P(x,y) = P(r.cosφ,r.sinφ)
P’(x’,y’)=P[r.cos(φ+θ),rsin(φ+θ)]

The coordinates of P’ are:

x’=r.cos(θ+φ)=r(cosθcosφ-sinθsinφ)
=x.cosθ-y.sinθ (where x=rcosφ and y=rsinφ)
similarly;
y’= rsin(θ+φ)=r(sinθcosφ + cosθ.sinφ)
=xsinθ+ycosθ

Thus,

Rθ = x’= x.cosθ-y.sinθ = R
θ
y’= xsinθ+ycosθ

Thus, we have obtained the new coordinate of point P after the rotation. In matrix
form, the transformation relation between P’ and P is given by:

cosθ sinθ
(x’y’)=(x,y)
-sinθ cosθ

that is P’=P.Rθ ---------(5)

where P’and P represent object points in 2-D Euclidean system and Rθ is

transformation matrix for anti-clockwise Rotation.

In terms of HCS, equation (5) becomes

cosθ sinθ 0
(x', y’,1) = (x, y, 1)
-sinθ cosθ 0 ---------(6)
0 0 1

That is P’h=Ph.Rθ , ---------(7)

10
 2-D and 3-D
Transformations
Where P’h and Ph represents object points, after and before required transformation, in
Homogeneous Coordinates and Rθ is called homogeneous transformation matrix for
anticlockwise Rotation. Thus, P’h, the new coordinates of a transformed object, can
be found by multiplying previous object coordinate matrix, Ph, with the transformation
matrix for Rotation Rθ.

Note that for clockwise rotation we have to put θ = – θ, thus the rotation matrix Rθ , in
HCS, becomes

cos(– θ) sin(–θ) 0 cosθ –sinθ 0

R-θ= –sin(–θ) cos(–θ) 0 = sinθ cosθ 0
0 0 1 0 0 1

Example 2: Perform a 450 rotation of a triangle A(0,0),B(1,1),C(5,2) about the origin.

Solution: We can represent the given triangle, in matrix form, using homogeneous
coordinates of the vertices:
A 0 0 1
[ABC]= B 1 1 1
C 5 2 1

cos450 sin450 0 √2/2 √2/2 0

0
The matrix of rotation is: Rθ = R45 = – sin450 cos450 0 = – √2/2 √2/2 0
0 0 1 0 0 1

So the new coordinates A’B’C’ of the rotated triangle ABC can be found as:

 0 0 1  2 / 2 2 / 2 0  0 0 1
   
[A’B’C’]=[ABC]. R45° = 1 1 1  − 2 / 2 2 / 2 0 =  0 2 1
 5 2 1  0 0

1  
  3 2 / 2 7 2 / 2 1

Thus A’=(0,0), B’=(0,√2), C’=(3√2/2,7√2/2)

The following Figure (a) shows the original, triangle [ABC] and Figure (b) shows
triangle after the rotation.

C
B

A x
O

Figure (a)

11
 Transformations
y

C’

B’

A’
x
O

Figure (b)

 Check Your Progress 1

1) What are the basic advantages of using Homogeneous coordinates system.
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………

2) A square consists of vertices A(0,0),B(0,1),C(1,1),D(1,0). After the translation C

is found to be at the new location (6,7). Determine the new location of other
vertices.
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………

3) A point P(3,3) makes a rotating of 450 about the origin and then translating in the
direction of vector v=5I+6J. Find the new location of P.
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………

4) Find the relationship between the rotations Rθ, R-θ, and Rθ-1 .
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………

1.2.3 Scaling

Scaling is the process of expanding or compressing the dimensions (i.e., size) of an

object. An important application of scaling is in the development of viewing
transformation, which is a mapping from a window used to clip the scene to a view
port for displaying the clipped scene on the screen.

12
 2-D and 3-D
Let P(x,y) be any point of a given object and sx and sy be scaling factors in x and y Transformations
directions respectively, then the coordinate of the scaled object can be obtained as:
x’=x.sx
y’=y.sy --------(8)

If the scale factor is 0<s<1, then it reduces the size of an object and if it is more then
1, it magnifies the size of the object along an axis.
For example, assume sx >1.

i) Consider (x,y)Æ(2.x,y) i.e., Magnification in x-direction with scale factor sx =2.

(3,3) (6,3)

(2,2) (4,2) sx =2 (4,2)

(2,1) (4,1) (4,1) (8,1)

Figure a): Object before Scaling Figure b): Object after Scaling with sx =2

ii) Similarly, assume sy >1 and consider (x,y)Æ(x,2.y), i.e., Magnification in y-

direction with scale factor sy =2.

(3,3) (3,6)

(2,2) (4,2) sy=2

(2,4) (4,4)

(2,1) (4,1)

(2,2) (4,2)

Figure a): Object before Scaling Figure b): Object after Scaling with Sy=2

iii) Consider (x,y)Æ(x.sx,y) where 0< sx = y2 <1 i.e., Compression in x-direction with
scale factor sx=1/2.

(3,3) (1.5,3)

(2,2) (4,2) sy =2

(2,1) (4,1) (1,2) (2,1)

Figure a): Object before Scaling Figure b): Object after Scaling with Sx=1/2

Thus, the general scaling is (x,y)Æ (x.sx,y.sy) i.e., magnifying or compression in both
x and y directions depending on Scale factors sx and sy. We can represent this in
matrix form (2-D Euclidean system) as:

sx 0
(x’,y’)= (x,y) 0 sy ----(9)

13
 Transformations
In terms of HCS, equation (9) becomes:

 sx 0 0
(x’,y’,1)=(x,y,1)  0 sy 0  ----(10)
 0 0 1 
that is P’h=Ph.ssx,sy ----(11)

Where Ph and P’h represents object points, before and after required transformation,
in Homogeneous Coordinates and ssx,sy is called transformation matrix for general
scaling with scaling factor sx and sy .

Thus, we have seen any positive value can be assigned to scale factors sx and sy. We
have the following three cases for scaling:

Case 1: If the values of sx and sy are less then 1, then the size of the object will be
reduced.

Case2: If both sx and sy are greater then 1, then the size of the object is enlarged.

Case3: If we have the same scaling factor (i.e. sx=sx=S), then there will be uniform
scaling (either enlargement or compression depending on the value of Sx and Sy) in
both x and y directions.

Example3: Find the new coordinates of a triangle A(0,0),B(1,1),C(5,2) after it has

been (a) magnified to twice its size and (b) reduced to half its size.

Solution: Magnification and reduction can be achieved by a uniform scaling of s

units in both the x and y directions. If, s>1, the scaling produces
magnification. If, s<1, the result is a reduction. The transformation can be
written as: (x,y,1)Æ(s.x,s.y,1). In matrix form this becomes

s 0 0
(x,y,1). 0 s 0 = (s.x,s.y,1)
0 0 1

We can represent the given triangle, shown in Figure (a), in matrix form, using
homogeneous coordinates of the vertices as :

A 0 0 1
B 1 1 1
C 5 2 1
y

5
4
3
C
2
B
1
A
x
O
1 2 3 4 5 6

Figure a: Object before scaling

14
 2-D and 3-D
(a) choosing s=2 Transformations
2 0 0
The matrix of scaling is:Ssx,sy = S2,2 = 0 2 0
0 0 1

So the new coordinates A’B’C’ of the scaled triangle ABC can be found as:

0 0 1 2 0 0 0 0 1
[A’B’C’]=[ABC]. R2,2 = 1 1 1 0 2 0 = 2 2 1
5 2 1 0 0 1 10 4 1

Thus, A’=(0,0), B’=(2,2), C’= (10, 4)

(b) Similarly, here, s=1/2 and the new coordinates are A’’=(0,0), B’’=(1/2,1/2),
C’’=(5/2,1). The following figure (b) shows the effect of scaling with sx=sy =2
and (c) with sx=sy =s=1/2.
y
y

4 4
C
3 B 3
2 B
C
2
1A
x 1A x
O O
1 2 3 4 5 6 7 8 9 10
1 2 3 4 5

Figure b: Object after scaling with Sx = Sy = 2 Figure c: Object after scaling with Sx = Sy = 1/2

1.2.4 Shearing

Shearing transformations are used for modifying the shapes of 2-D or 3-D objects.
The effect of a shear transformation looks like “pushing” a geometric object in a
direction that is parallel to a coordinate plane (3D) or a coordinate axis (2D). How far
a direction is pushed is determined by its shearing factor.

One familiar example of shear is that observed when the top of a book is moved
relative to the bottom which is fixed on the table.

In case of 2-D shearing, we have two types namely x-shear and y-shear.

In x-shear, one can push in the x-direction, positive or negative, and keep the y-
direction unchanged, while in y-shear, one can push in the y-direction and keep the x-
direction fixed.

x-shear about the origin

Let an object point P(x,y) be moved to P’(x’,y’) in the x-direction, by the given scale
parameter ‘a’,i.e., P’(x’y’) be the result of x-shear of point P(x,y) by scale factor a
about the origin, which is shown in Figure 4.

15
 Transformations
ay
y
ay
P(x,y) P(x,y)
P’ (x’,y’)

x
O O x
x

Figure 4

Thus, the points P(x,y) and P’(x’,y’) have the following relationship:

x' = x + ay
y' = y = Shx(a) ------(11a)

where ‘a’ is a constant (known as shear parameter) that measures the degree of
shearing. If a is negative then the shearing is in the opposite direction.

Note that P(0,H) is taken into P'(aH,H). It follows that the shearing angle A (the angle
through which the vertical edge was sheared) is given by:

tan(A) = aH/H = a.

So the parameter a is just the tan of the shearing angle. In matrix form (2-D Euclidean
system), we have

1 0
(x’,y’)=(x,y) a 1 -------(12)

In terms of Homogeneous Coordinates, equation (12) becomes

1 0 0
(x’,y’,1)=(x,y,1). a 1 0 -------(13)
0 0 1

That is, P’h = Ph Shx(a) --------(14)

Where Ph and P’h represents object points, before and after required transformation,
in Homogeneous Coordinates and Shx(a) is called homogeneous transformation matrix
for x-shear with scale parameter ‘a’ in the x-direction.

y-shear about the origin

Let an object point P(x,y) be moved to P’(x’,y’) in the x-direction, by the given scale
parameter ‘b’. i.e., P’(x’y’) be the result of y-shear of point P(x,y) by scale factor ‘b’
about the origin, which is shown in Figure 5(a).

Y Y P’ (x’,y’)

P (x,y) b.x
P(x, y)

X
X

Figure 5 (a)

16
 2-D and 3-D
Thus, the points P(x,y) and P’(x’,y’) have the following relationship : Transformations

x' = x
y' = y+bx = Shy(b) ----------(15)

where ‘b’ is a constant (known as shear parameter) that measures the degree of
shearing. In matrix form, we have

1 b
(x’,y’)=(x,y) 0 1 ----------(16)

In terms of Homogeneous Coordinates, equation (16) becomes

1 b 0
(x’,y’,1)=(x,y,1) 0 1 0 ---------(17)
0 0 1

That is, P’h = Ph.Shy(b) ---------(18)

Where Ph and P’h represents object points, before and after required transformation,
in Homogeneous Coordinates and Shy (b) is called homogeneous transformation
matrix for y-shear with scale factor ‘b’ in the y-direction.

xy-shear about the origin

Let an object point P(x,y) be moved to P’(x’,y’) as a result of shear transformation in

both x- and y-directions with shearing factors a and b, respectively, as shown in
Figure 5(b). a.y
Y b.x
P’ (x’, y’)

a.y
P (x, y)

b.x

O X
Figure 5 (b)

The points P(x,y) and P’(x’,y’) have the following relationship :

x' = x +ay
y' = y+bx = Shxy(a,b) ----------(19)

where ′ay′ and ′bx′ are shear factors in x and y directions, respectively. The xy-shear
is also called simultaneous shearing or shearing for short.

In matrix form, we have,

1 b
(x’,y’)=(x,y) a 1 ---------(20)

In terms of Homogeneous Coordinates, we have

1 b 0
(x’,y’,1)=(x,y,1) a 1 0 ---------(21)
0 0 1

That is, P’h = Ph.Shxy(a,b) ----------(22)

17
 Transformations
Where Ph and P’h represent object points, before and after required transformation, in
Homogeneous Coordinates and Shxy(a,b) is called homogeneous transformation matrix
for xy-shear in both x- and y-directions with shearing factors a and b, respectively,

Special case: when we put b=0 in equation (21), we have shearing in x-direction, and
when a=0, we have Shearing in the y-direction, respectively.

Example 4: A square ABCD is given with vertices A(0,0),B(1,0),C(1,1), and D(0,1).

Illustrate the effect of a) x-shear b) y-shear c) xy-shear on the given square, when a=2
and b=3.

Solution: We can represent the given square ABCD, in matrix form, using
homogeneous coordinates of vertices as:

A 0 0 1
B 1 0 1
C 1 1 1
D 0 1 1

a) The matrix of x-shear is:

1 0 0 1 0 0
Shx(a) = a 1 0 = 2 1 0
0 0 1 0 0 1

So the new coordinates A’B’C’D’ of the x-sheared object ABCD can be found as:
[A’B’C’D’]=[ABCD]. Shx(a)

A 0 0 1 1 0 0 0 0 1
[A’B’C’D’]= B 1 0 1 . 2 1 0 = 1 0 1
C 1 1 1 0 0 1 3 1 1
D 0 1 1 2 1 1

Thus, A’=(0,0), B’=(1,0), C’=(3,1) and D’=(2,1).

b) Similarly the effect of shearing in the y direction can be found as:

[A’B’C’D’]=[ABCD].Shy(b)

A 0 0 1 1 3 0 0 0 1
[A’B’C’D’]= B 1 0 1 . 0 1 0 = 1 3 1
C 1 1 1 0 0 1 1 4 1
D 0 1 1 0 1 1

Thus, A’=(0,0), B’=(1,3), C’=(1,4) and D’=(0,1).

c) Finally the effect of shearing in both directions can be found as:

[A’B’C’D’]=[ABCD]. Shxy(a,b)

A 0 0 1 1 3 0 0 0 1
[A’B’C’D’]= B 1 0 1 . 2 1 0 = 1 3 1
C 1 1 1 0 0 1 3 4 1
D 0 1 1 2 1 1

Thus, A’=(0,0), B’=(1,3), C’=(3,4) and D’=(2,1).

18
 2-D and 3-D
Transformations
Figure (a) shows the original square, figure (b)-(d) shows shearing in the x, y and
both directions respectively.
Y
Y

C (1,1)
D D’ (2, 1)
D C’ (3,1)
A B
O A
X B X
O

Figure (a) Figure (b)

Y Y

C’ (1, 4)
C’ (3,4)
D’ (1, 3)
B’ (1, 3)
D’ D
B’(2, 1)
A A’
O O
X

Figure (c) Figure (d)

Example 5: What is the use of Inverse transformation? Give the Inverse

transformation for translation, rotation, reflection, scaling, and shearing.

Solution: We have seen the basic matrix transformations for translation, rotation,
reflection, scaling and shearing with respect to the origin of the coordinate system. By
multiplying these basic matrix transformations, we can build complex
transformations, such as rotation about an arbitrary point, mirror reflection about a
line etc. This process is called concatenation of matrices and the resulting matrix is
often referred to as the composite transformation matrix. Inverse transformations play
an important role when you are dealing with composite transformation. They come to
the rescue of basic transformations by making them applicable during the construction
of composite transformation. You can observed that the Inverse transformations for
translation, rotation, reflection, scaling and shearing have the following relations, and
v, θ, a, b, sx, sy, sz are all parameter involved in the transformations.

1) Tv –1 =T-v
2) Rθ –1 = R-θ
3) (i) Shx-1(a) =Shx(-a)
(ii) Shy-1(b) =Shx(-b)
(iii) Shxy-1(a,b) =Shx(-a,-b)
4) S-1sx,sy,sz =S1/sx,1/sy,1/sz
5) The transformation for mirror reflection about principal axes do not change after
inversion.
(i) Mx-1 =M-x= Mx
(ii) My-1 =M-y= My
(iii) Mz-1 =M-z= Mz ,

19
 Transformations
6) The transformation for rotations made about x,y,z axes have the following
inverse:
(i) R-1x,θ = Rx,-θ = RTx,θ
(ii) R-1y,θ = Ry,-θ = RTy,θ
(iii) R-1z,θ = Rz,-θ = RTz,θ

 Check Your Progress 2

1) Differentiate between the Scaling and Shearing transformation.
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………

2) Show that Sa,b .Sc,d = Sc,d .Sa,b = Sac,bd

………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………

3) Find the 3x3 homogeneous co-ordinate transformation matrix for each of the
following:
a) Shift an image to the right by 3 units.
b) Shift the image up by 2 units and down 1 units.
c) Move the image down 2/3 units and left 4 units.
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………

4) Find the condition under which we have Ssx,sy .Rθ= Rθ. Ssx,sy.
………………………………………………………………………………………
………………………………………………………………………………………
…………………………….………………………………………………………

5) Is a simultaneous shearing the same as the shearing in one direction followed by a

shearing in another direction? Why?
………………………………………………………………………………………
………………………………………………………………………………………
…………………………….………………………………………………………

1.3 COMPOSITE TRANSFORMATIONS

We can build complex transformations such as rotation about an arbitrary point,

mirror reflection about a line, etc., by multiplying the basic matrix transformations.
This process is called concatenation of matrices and the resulting matrix is often
referred to as the composite transformation matrix. In composite transformation, a
previous transformation is pre-multiplied with the next one.

20
 2-D and 3-D
In other words we can say that a sequence of the transformation matrices can be Transformations
concatenated into a single matrix. This is an effective procedure as it reduces because
instead of applying initial coordinate position of an object to each transformation
matrix, we can obtain the final transformed position of an object by applying
composite matrix to the initial coordinate position of an object. In other words we can
say that a sequence of transformation matrix can be concatenated matrix into a single
matrix. This is an effective procedure as it reduces computation because instead of
applying initial coordinate position of an object to each transformation matrix, we can
obtain the final transformed position of an object by applying composite matrix to the
initial coordinate position of an object.

1.3.1 Rotation about a Point

Given a 2-D point P(x,y), which we want to rotate, with respect to an arbitrary point
A(h,k). Let P’(x’y’) be the result of anticlockwise rotation of point P by angle θ about
A, which is shown in Figure 6.

P’(x’, y’)
Y’
Y P(x, y)
θ
φ
A(h, k) X’
O
X
Figure 6

Since, the rotation matrix Rθ is defined only with respect to the origin, we need a set
of basic transformations, which constitutes the composite transformation to compute
the rotation about a given arbitrary point A, denoted by Rθ,A. We can determine the
transformation Rθ,A in three steps:

1) Translate the point A(h,k) to the origin O, so that the center of rotation A is at the
origin.
2) Perform the required rotation of θ degrees about the origin, and
3) Translate the origin back to the original position A(h,k).

Using v=hI+kJ as the translation vector, we have the following sequence of three
transformations:

Rθ,A =T-v. Rθ. Tv

1 0 0 cosθ sinθ 0 1 0 0
= 0 1 0 – sinθ cosθ 0 0 1 0
–h – k 1 0 0 1 h k 1

cosθ sinθ 0
= – sinθ cosθ 0 -------(23)
(1– cosθ).h+k.sinθ (1– cosθ).k – h.sinθ 1

Example 5: Perform a 450 rotation of a triangle A (0,0), B (1,1), C (5,2) about an

arbitrary point P(–1, –1).

Solution: Given triangle ABC, as show in Figure (a), can be represented in

homogeneous coordinates of vertices as:

21
 Transformations
A  x1 y1 1   0 0 1
   
[A B C ] = B  x2 y2 1  =  1 1 1
   5 2 1
C  x3 y3 1   

From equation (23), a rotation matrix RQ, A about a given arbitrary point A (h, k) is:

 Cos θ Sinθ 0
 
Rq , A =  − Sin θ Cos θ 0
 (1 − Cosθ ). h + k . Sinθ (1 − Cosθ ). k − h . Sinθ 1 
 

 2 /2 2 / 2 0 

Thus R45 o , A=− 2 / 2 2 / 2 0


−1 ( 2 −1 1  )



So the new coordinates [A′ B′ C ′] of the rotated triangle [ABC] can be found as:

 0 0 1  2 /2 2 / 2 0 
  
[A′ B′ C ′] = [ABC] . R45 o , A =  1 1 1 . − 2 / 2 2 / 2 0 =
 5 2 1
 


−1 ( )
2 −1 1 


 
A′ 

−1 2 −1 ( 1

)
B′  −1 2 2 − 1 1
C ′  3 2 −1  9 . 2 −1 1
 2  2 
  

Thus, A′ = (− 1, (
2 −1 ), B′ = − 1, 2 2 −1 , and C ′ = 
3
2 − 1,)9 
2 − 1 . The
2 2 
following figure (a) and (b) shows a given triangle, before and after the rotation.
Y
y
C’
6
5
5
4
C (5, 2) 4
3
B (1, 1) 3
2
B’
2
1
A’ 1
O
X X
1 2 3 4 5 6 1 2 3 4 5 6
Figure (a) Figure (b)

1.3.2 Reflection about a Line

Reflection is a transformation which generates the mirror image of an object. As

discussed in the previous block, the mirror reflection helps in achieving 8-way
symmetry for the circle to simplify the scan conversion process. For reflection we
need to know the reference axis or reference plane depending on whether the object is
2-D or 3-D.

22
 2-D and 3-D
Transformations
Let the line L be represented by y=mx+c, where ‘m’ is the slope with respect to the x
axis, and ‘c’ is the intercept on y-axis, as shown in Figure 7. Let P’(x’,y’) be the
mirror reflection about the line L of point P(x,y).
Y
L
Y = mx + c

(O, C)

L′
θ
θ
X

Figure 7

The transformation about mirror reflection about this line L consists of the following
basic transformations:

1) Translate the intersection point A(0,c) to the origin, this shifts the line L to L’.
2) Rotate the shifted line L’ by –θ degrees so that the line L’ aligns with the x-axis.
3) Mirror reflection about x-axis.
4) Rotate the x-axis back by θ degrees
5) Translate the origin back to the intercept point (0,c).

In transformation notation, we have

ML= T-v.R-θ.MX.Rθ.Tv , where v=0I+cJ

1 0 0 cosθ –sinθ 0 1 0 0 cosθ sinθ 0 1 0 0

ML = 0 1 0 sinθ cosθ 0 0 –1 0 –sinθ cosθ 0 0 1 0
0 –c 1 0 0 1 0 0 1 0 0 1 0 c 1

cos2θ – sin2θ 2.cosθ.sinθ 0

= 2.sinθ.cosθ sin2θ – cos2θ 0 ----------(24)
2 2
–2.c.sinθ.cosθ – c.(sin θ – cos θ)+c 1

Let tanθ=m, the standard trigonometry yields sinθ=m/√(m2+1) and cosθ= 1/√(m2+1).
Substituting these values for sinθ and cosθ in the equation (24), we have:

ML= (1– m2)/ (m2 +1) 2m/(m2 +1) 0

2m/(m2 +1) (m2 –1)/(m2 +1) 0 ------------(25)
2
–2cm/(m +1) 2c/(m2 +1) 1

Special cases

1) If we put c = 0 and m=tanθ=0 in the equation (25) then we have the reflection
about the line y = 0 i.e. about x-axis. In matrix form:

1 0 0
Mx= 0 –1 0 ----------(26)
0 0 1

2) If c = 0 and m=tanθ=∞ then we have the reflection about the line x=0 i.e. about
y-axis. In matrix form:
–1 0 0
My= 0 1 0 ---------(27)
0 0 1
23
 Transformations
4) To get the mirror reflection about the line y = x, we have to put m=1 and c=0.
In matrix form:
0 1 0
My=x = 1 0 0 -----------(28)
0 0 1

5) Similarly, to get the mirror reflection about the line y = – x, we have to put m = –1
and c = 0. In matrix form:
0 –1 0
My=-x = –1 0 0 -----------(29)
0 0 1

6) The mirror reflection about the Origin (i.e., an axis perpendicular to the xy plane
and passing through the origin).
–1 0 0
Morg = 0 –1 0 ----------(30)
0 0 1
y

z x

mirror
Figure 7(a)

Example 6: Show that two successive reflections about either of the coordinate axes
is equivalent to a single rotation about the coordinate origin.

Solution: Let (x, y) be any object point, as shown in Figure (a). Two successive
reflection of P, either of the coordinate axes, i.e., Reflection about x-axis followed by
reflection about y-axis or vice-versa can be reprosecuted as:

(x, y) Mx
(x, − y) My (−x, −y) ----(i) E

A D
.P(x, y)

(x, y) My (x, − y) Mx (−x, −y) ----(ii) B C

Figure (a)

The effect of (1) and (2) can also be illustrated by the following Figure (b) and
E
Figure (c)

A D
Reflection about x-axis
P

B
C

B’ C’
C” B”

P” P’
A’
D” A” D’

E’
E” Reflection about y-axis
24 Figure (b)
 2-D and 3-D
Transformations

Reflection about y-axis

E’ E

A’
D’ A D

P’ P

C’ B’ B
C

P”
Figure (c)

Reflection about x-axis

From equation (i) and (ii), we can write:
−1 0 
(x, y) (− x, − y) = (x, y)   (iii)
 0 − 1
Equation (3) is the required reflection about the origin. Hence, two successive
reflections about either of the coordinate axes is just equivalent to a single rotation
about the coordinate origin.

Example 7: Find the transforation matrix for the reflection about the line y = x.

Solution: The transformation for mirror reflection about the line y = x, consists of the
following three basic transformations.

y=x
o
45

Figure (a)

1) Rotate the line L through 45o in clockwise rotation,

2) Perform the required Reflection about the x-axis.
3) Rotate back the line L by − 45o
i.e.,

ML = R 45o . Mx . R -45o

Cos 45o − Sin 45o 0 1 0 0   Cos 45o + Sin 45o 0

  0 − 1 0  .  
=  Sin 45o Cos 45o 0 .  
o
 − Sin 45 Cos 45
o
0
 0 0 1 0 0 1  0 0 1 
  

25
 Transformations
Cos 45o Sin 45o 0  Cos 45o Sin 45o 0
   
=  Sin 45o − Cos 45o 0 . − Sin 45o Cos 45o 0
 0 0 1  0 0 1
 
Cos 90 Sin 90 0 1 0 0 
 
=  Sin 90 − Cos 90 0 =  0 0 0  = My = x
 0 0 1  0 0 1 

Example 8 : Reflect the diamond-shaped polygon whose vertices are A(–1,0),

B(0, –2),C(1,0) and D(0,2) about (a) the horizontal line y=2, (b) the vertical line x=2,
and (c) the line y=x+2.

Solution: We can represent the given polygon by the homogeneous coordinate matrix
as
–1 0 1
V=[ABCD] = 0 –2 1
1 0 1
0 2 1
a) The horizontal line y=2 has an intercept (0,2) on y axis and makes an angle of 0
degree with the x axis. So m=0 and c=2. Thus, the reflection matrix

ML= T-v.R-θ.Mx.Rθ.T-v , where v=0I+2J

1 0 0
= 0 –1 0
0 4 1
So the new coordinates A’B’C’D’ of the reflected polygon ABCD can be found
as:
[A’B’C’D’]=[ABCD]. ML

–1 0 1 1 0 0 –1 4 1
= 0 –2 1 . 0 –1 0 = 0 6 1
1 0 1 0 4 1 1 4 1
0 2 1 0 2 1

Thus, A’=(-1,4), B’=(0,6), C’=(1,4) and D’=(0,2).

b) The vertical line x=2 has no intercept on y-axis and makes an angle of 90 degree
with the x-axis. So m=tan900=∞ and c=0. Thus, the reflection matrix

ML= T-v.R-θ.My.Rθ.T-v , where v=2I

–1 0 0
= 0 1 0
4 0 1

So the new coordinates A’B’C’D’ of the reflected polygon ABCD can be found
as:
[A’B’C’D’]=[ABCD]. ML

–1 0 1 –1 0 0 5 0 1
= 0 –2 1 . 0 1 0 = 4 –2 1
1 0 1 4 0 1 3 0 1
0 2 1 4 2 1

Thus, A’=(5,0), B’=(4,-2), C’=(3,0) and D’=(4,2)

26
 2-D and 3-D
c) The line y=x+2 has an intercept (0,2) on y-axis and makes an angle of 450 with Transformations
the x-axis. So m=tan450=1 and c=2. Thus, the reflection matrix

0 1 0
ML= 1 0 0
–2 2 1

The required coordinates A’,B’, C’, and D’ can be found as:

[A’B’C’D’]=[ABCD]. ML

 −1 0 1  −2 1 1
   0 1 0  
 0 −2 1    −4 2 1
1 0 1  1 0 0 . =
 −2 3 1
   −2 2 1   
 
 0 2 1  0 2 1

Thus, A’=(–2,1), B’=(–4,2), C’=(–2,3) and D’=(0,2)

The effect of the reflected polygon, which is shown in Figure (a), about the line y=2,
x=2, and y=x+2 is shown in Figure (b) - (d), respectively.
B’

A’ C’
D

D’
A C
O O

Figure (a) Figure (b)

D’ C’

B’
D’
C’ A’
A’
O O

B’

Figure (c) Figure (d)

1.4 HOMOGENEOUS COORDINATE SYSTEMS

Let P(x,y) be any point in 2-D Euclidean (Cartesian) system.

In Homogeneous Coordinate system, we add a third coordinate to a point. Instead of

(x,y), each point is represented by a triple (x,y,H) such that H≠0; with the condition
that (x1,y1,H1)=(x2,y2,H2) ↔ x1/H1 = x2/H2 ; y1/H1 = y2/H2.

(Here, if we take H=0, then we have point at infinity, i.e., generation of horizons).

27
 Transformations
Thus, (2,3,6) and (4,6,12) are the same points are represented by different coordinate
triples, i.e., each point has many different Homogeneous Coordinate representation.

2-D Euclidian System Homogeneous Coordinate System

Any point (x,y) (x,y,1)

If (x,y,H) be any point in HCS(such that H≠0);

Then (x,y,H)=(x/H,y/H,1)

(x/H,y/H) (x,y,H)

Now, we are in the position to construct the matrix form for the translation with the
use of homogeneous coordinates.

For translation transformation (x,y)Æ(x+tx,y+ty) in Euclidian system, where tx and ty

are the translation factor in x and y direction, respectively. Unfortunately, this way of
describing translation does not use a matrix, so it cannot be combined with other
transformations by simple matrix multiplication. Such a combination would be
desirable; for example, we have seen that rotation about an arbitrary point can be done
by a translation, a rotation, and another translation. We would like to be able to
combine these three transformations into a single transformation for the sake of
efficiency and elegance. One way of doing this is to use homogeneous coordinates. In
homogeneous coordinates we use 3x3 matrices instead of 2x2, introducing an
additional dummy coordinate H. Instead of (x,y), each point is represented by a triple
(x,y,H) such that H≠0; In two dimensions the value of H is usually kept at 1 for
simplicity.

Thus, in HCS (x,y,1) → (x+tx,y+ty,1), now, we can express this in matrix form as:

1 0 0
(x’,y’,1)=(x,y,1) 0 1 0
tx ty 1

The advantage of introducing the matrix form of translation is that it simplifies the
operations on complex objects, i.e., we can now build complex transformations by
multiplying the basic matrix transformations.

In other words, we can say, that a sequence of transformation matrices can be

concatenated into a single matrix. This is an effective procedure as it reduces the
computation because instead of applying initial coordinate position of an object to
each transformation matrix, we can obtain the final transformed position of an object
by applying composite matrix to the initial coordinate position of an object. Matrix
representation is standard method of implementing transformations in computer
graphics.

Thus, from the point of view of matrix multiplication, with the matrix of translation,
the other basic transformations such as scaling, rotation, reflection, etc., can also be
expressed as 3x3 homogeneous coordinate matrices. This can be accomplished by
augmenting the 2x2 matrices with a third row (0,0,x) and a third column. That is

0 a b 0
0 c d 0
1 0 0 1

28
 2-D and 3-D
Example 9: Show that the order in which transformations are performed is important Transformations
by applying the transformation of the triangle ABC by:

(i) Rotating by 45o about the origin and then translating in the direction of the vector
(1,0), and

(ii) Translating first in the direction of the vector (1,0), and then rotating by 45o about
the origin, where A = (1, 0) B = (0 ,1) and C = (1, 1).

Solution: We can represent the given triangle, as shown in Figure (a), in terms of
Homogeneous coordinates as:

1 0 1 B C (1,1)

[ABC] = 0 1 1

A
1 1 1

Figure (a)

Suppose the rotation is made in the counter clockwise direction. Then, the
transformation matrix for rotation, R 45o , in terms of homogeneous coordinate system
is given by:
 Cos 45o Sin 45o 0  1 / 2 1 / 2 0
   
R 45o = − Sin 45o Cos 45o 0 =  − 1 / 2 1 / 2 0
 0 0 1  0 0 1
  

and the Translation matrix, Tv, where V = 1I + 0J is:

 1 0 0 1 0 0 
 
Tv =  0 1 0 = 0 1 0
 t x t y 1 1 0 0
 

where tx and ty is the translation factors in the x and y directions respectively.

i) Now the rotation followed by translation can be computed as:

 1 / 2 1 / 2 0 1 0 0  1 / 2 1 / 2 0
   
R 45o . Tv = − 1 / 2 1 / 2 0 . 0 1 0 = − 1 / 2 1 / 2 0
 0 0 1 1 0 1  1 0 1
   

So the new coordinates A ′ B′ C ′ of a given triangle ABC can be found as:

[A ′ B′ C′] = [A B C] . R 45o . Tv

1 0 1  1 / 2 1 / 2 0 (
 1 / 2 +1 ) 1/ 2 1
 
= 0 1 1 . − 1 / 2 1 / 2 0 =

(
 −1 / 2 + 1 ) 1/ 2

1 (I)
1 1 1  1 0 1  1 2 1
   

implies that the given triangle A(1,0), B (0, 1) C (1, 1) be transformed into
 1
A ′  + 1,
1   −1
, B′  + 1,
1 
( )
 and C ′ 1, 2 , respectively, as shown in
 2 2  2 2
Figure (b).

29
 Transformations

C’

B’ A’

Figure (b)

Similarly, we can obtain the translation followed by rotation transformation as:

1 0 0  1 / 2 1 / 2 0  1 / 2 1 / 2 0
Tv . R 45o 0 1 0  .  − 1 / 2 1 / 2 0  =  − 1 / 2 1 / 2 0 
     
1 0 1  0 0 1  1 / 2 1 / 2 1
   

And hence, the new coordinates A ′ B′ C ′ can be computed as:

[A ′ B′ C′] = [A B C]. Tv R 45o

1 0 1  1 / 2 1 / 2 0 2 / 2 2 / 2 1
= 0 1 1 . − 1 / 2 1 / 2 0 =  0 
2 / 2 1 (II)
    
1 1 1  1 / 2 1 / 2 1 1 / 2 3 / 2 1
   

Thus, in this case, the given triangle A(1,0), B(0, 1) and C(1,1) are transformed into

( ) ( ) 1
A′′ 2 / 2, 2 / 2 , B ′′ 0, 2 / 2 and C ′′  ,
3 
, respectively, as shown in
 2 2
Figure (c).

C’’
B’’ A’’

Figure (c)

By (I) and (II), we see that the two transformations do not commute.

 Check Your Progress 3

1) Show that transformation matrix (28), for the reflection about the line y=x, is
equivalent to the reflection relative to the x-axis followed by a counterclockwise
rotation of 900.
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………

30
 2-D and 3-D
2) Give a single 3x3 homogeneous coordinate transformation matrix, which will Transformations
have the same effect as each of the following transformation sequences.
a) Scale the image to be twice as large and then translate it 1 unit to the left.
b) Scale the x direction to be one-half as large and then rotate counterclockwise
by 900 about the origin.
c) Rotate counterclockwise about the origin by 900 and then scale the x direction
to be one-half as large.
d) Translate down ½ unit, right ½ unit, and then rotate counterclockwise by 450.

3) Obtain the transformation matrix for mirror reflection with respect to the line
y=ax, where ‘a’ is a constant.
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………

4) Obtain the mirror reflection of the triangle formed by the vertices A(0,3),B(2,0)
and C(3,2) about the line passing through the points (1,3) and (–1, –1).
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………

1.5 3-D TRANSFORMATIONS

The ability to represent or display a three-dimensional object is fundamental to the

understanding of the shape of that object. Furthermore, the ability to rotate, translate,
and project views of that object is also, in many cases, fundamental to the
understanding of its shape. Manipulation, viewing, and construction of three-
dimensional graphic images require the use of three-dimensional geometric and
coordinate transformations. In geometric transformation, the coordinate system is
fixed, and the desired transformation of the object is done with respect to the
coordinate system. In coordinate transformation, the object is fixed and the desired
transformation of the object is done on the coordinate system itself. These
transformations are formed by composing the basic transformations of translation,
scaling, and rotation. Each of these transformations can be represented as a matrix
transformation. This permits more complex transformations to be built up by use of
matrix multiplication or concatenation. We can construct the complex
objects/pictures, by instant transformations. In order to represent all these
transformations, we need to use homogeneous coordinates.

Hence, if P(x,y,z) be any point in 3-D space, then in HCS, we add a fourth-coordinate
to a point. That is instead of (x,y,z), each point can be represented by a Quadruple
(x,y,z,H) such that H≠0; with the condition that x1/H1=x2/H2; y1/H1=y2/H2;
z1/H1=z2/H2. For two points (x1, y1, z1, H1) = (x2, y2, z2, H2) where H1 ≠ 0, H2 ≠ 0.
Thus any point (x,y,z) in Cartesian system can be represented by a four-dimensional
vector as (x,y,z,1) in HCS. Similarly, if (x,y,z,H) be any point in HCS then
(x/H,y/H,z/H) be the corresponding point in Cartesian system. Thus, a point in three-
dimensional space (x,y,z) can be represented by a four-dimensional point as:
(x’,y’,z’,1)=(x,y,z,1).[T], where [T] is some transformation matrix and (x’,y’z’,1) is a
new coordinate of a given point (x,y,z,1), after the transformation.

31
 Transformations
The generalized 4x4 transformation matrix for three-dimensional homogeneous
coordinates is:

a b c w
d e f x (3x3) (3x1)
[T]= g h I y = --------(31)
l m n z
(1x3) (1x1)

The upper left (3x3) sub matrix produces scaling, shearing, rotation and reflection
transformation. The lower left (1x3) sub matrix produces translation, and the upper
right (3x1) sub matrix produces a perspective transformation, which we will study in
the next unit. The final lower right-hand (1x1) sub matrix produces overall scaling.

1.5.1 Transformation for 3-D Translation

Let P be the point object with the coordinate (x,y,z). We wish to translate this object
point to the new position say, P’(x’,y’,z’) by the translation Vector V=tx.I+ty.J+tz.K ,
where tx , ty and tz are the translation factor in the x, y, and z directions respectively, as
shown in Figure 8. That is, a point (x,y,z) is moved to (x+ tx,y+ ty,z+ tz). Thus the new
coordinates of a point can be written as:
x’=x+ tx
y’=y+ty =Tv ---------(32)
z’=z+tz
z
(x + tx, y + ty, z + tz)
P’ (x’, y’, z’)
v'
v
P (x, y, z)

x Figure 8

In terms of homogeneous coordinates, equation (32) can be written as

1 0 0 0
(x’,y’,z’,1)=(x,y,z,1) 0 1 0 0 --------(33)
0 0 1 0
tx ty tz 1

i.e., P’h = Ph.Tv ----------(34)

1.5.2 Transformation for 3-D Rotation

Rotation in three dimensions is considerably more complex then rotation in two

dimensions. In 2-D, a rotation is prescribed by an angle of rotation θ and a centre of
rotation, say P.

However, in 3-D rotations, we need to mention the angle of rotation and the axis of
rotation. Since, we have now three axes, so the rotation can take place about any one
of these axes. Thus, we have rotation about x-axis, y-axis, and z-axis respectively.

32
 2-D and 3-D
Rotation about z-axis Transformations

Rotation about z-axis is defined by the xy-plane. Let a 3-D point P(x,y,z) be rotated to
P’(x’,y’,z’) with angle of rotation θ see Figure 9. Since both P and P’ lies on xy-plane
i.e., z=0 plane their z components remains the same, that is z=z’=0.
Z
Z

Y
P’(x’, y’, z’) Y
θ
θ P’(x’, y’, o)
X
P(x, y, z) X
P(x, y, o)

Figure 9 Figure 10

Thus, P’(x’y’,0) be the result of rotation of point P(x,y,0) making a positive

(anticlockwise) angle φ with respect to z=0 plane, as shown in Figure 10.

From figure (10),

P(x,y,0) = P(r.cosφ,r.sinφ,0)
P’(x’,y’,0)=P[r.cos(φ+θ),rsin(φ+θ),0]

The coordinates of P’ are:

x’=r.cos(θ+φ)=r(cosθcosφ – sinθsinφ)
=x.cosθ – y.sinθ (where x=rcosφ and y=rsinφ)

similarly;

y’= rsin(θ+φ)=r(sinθcosφ + cosθ.sinφ)

=xsinθ+ycosθ
Thus,
x’= x.cosθ – y.sinθ
[Rz]θ = y’= xsinθ + ycosθ ----------(35)
z’=z

In matrix form,

cosθ sinθ 0
(x’y’,z’)=(x,y,z) –sinθ cosθ 0 ----------(36)
0 0 1

In terms of HCS, equation (36) becomes

cosθ sinθ 0 0
(x’y’,z’,1)=(x,y,z,1) –sinθ cosθ 0 0 ----------(37)
0 0 1 0
0 0 0 1
That is, P’h = Ph.[Rz]θ ---------(38)

33
 Transformations
Rotations about x-axis and y-axis

Rotation about the x-axis can be obtained by cyclic interchange of xÆyÆzÆx in

equation (35) of the z-axis rotation i.e.,

x’= x.cosθ-y.sinθ
[Rz]θ= y’= xsinθ+ycosθ
z’=z

After cyclic interchange of xÆyÆzÆx

y’= y.cosθ-z.sinθ
[Rx]θ= z’= y.sinθ+z.cosθ --------(39)
x’= x

So, the corresponding transformation matrix in homogeneous coordinates becomes

1 0 0 0
(x’y’,z’,1)=(x,y,z,1) 0 cosθ sinθ 0
0 – sinθ cosθ 0
0 0 0 1

That is, P’h = Ph.[Rx]θ -------(40)

Similarly, the rotation about y-axis can be obtained by cyclic interchange of

xÆyÆzÆx in equation (39) of the x-axis rotation [Rx]θ i.e.,

y’= y.cosθ-z.sinθ
[Rx]θ= z’= y.sinθ+z.cosθ
x’= x

After cyclic interchange of xÆyÆzÆx

z’= z.cosθ-x.sinθ
[Ry]θ= x’= z.sinθ+x.cosθ -------(41)
y’= y

So, the corresponding transformation matrix in homogeneous coordinates becomes

cosθ 0 – sinθ 0
(x’y’,z’,1)=(x,y,z,1) 0 1 0 0
sinθ 0 cosθ 0
0 0 0 1

That is, P’=P. [Ry]θ -------(42)

1.5.3 Transformation for 3-D Scaling

As we have seen earlier, the scaling process is mainly used to change the size of an
object. The scale factors determine whether the scaling is a magnification, s>1, or a

34
 2-D and 3-D
reduction, s<1. Two-dimensional scaling, as in equation (8), can be easily extended to Transformations
scaling in 3-D case by including the z-dimension.

For any point (x,y,z), we move into (x.sx,y.sy,z.sz), where sx, sy, and sz are the scaling
factors in the x,y, and z-directions respectively.

Thus, scaling with respect to origin is given by:

x’= x.sx
Ssx,sy,sz = y’= y.sy ---------(43)
z’= z.sz

In matrix form,

sx 0 0
(x’y’,z’)=(x,y,z) 0 sy 0 ---------(44)
0 0 sz

In terms of HCS, equation (44) becomes

sx 0 0 0
(x’y’,z’,1)=(x,y,z,1) 0 sy 0 0
0 0 sz 0
0 0 0 1

That is, P’=P. Ssx,sy,sz --------(45)

1.5.4 Transformation for 3-D Shearing

Two-dimensional xy- shearing transformation, as defined in equation (19), can also be

easily extended to 3-D case. Each coordinate is translated as a function of
displacements of the other two coordinates. That is,

x’=x+a.y+b.z
Shxyz= y’=y+c.x+d.z --------(46)
z’=z+e.x+f.y

where a,b,c,d,e and f are the shearing factors in the respective directions.

In terms of HCS, equation (46) becomes

1 c e 0
(x’y’,z’,1)=(x,y,z,1) a 1 f 0
b d 1 0
0 0 0 1

That is, P’h = Ph.Shxyz -------(47)

Note that the off-diagonal terms in the upper left 3x3 sub matrix of the generalized
4x4 transformation matrix in equation (31) produce shear in three dimensions.

1.5.5 Transformation for 3-D Reflection

For 3-D reflections, we need to know the reference plane, i.e., a plane about which the
reflection is to be taken. Note that for each reference plane, the points lying on the
plane will remain the same after the reflection.

35
 Transformations
Mirror reflection about xy-plane

Let P(x,y,z) be the object point, whose mirror reflection is to be obtained about xy-
plane(or z=0 plane). For the mirror reflection of P about xy-plane, only there is a
change in the sign of z-coordinate, as shown in Figure (11). That is,
Z
x’=x
Mxy = y’=y -------(48)
P(x, y, z)
z’=-z

In matrix form, Y

Z = 0 plane

X
P P(x, y, −z)

Figure 11

1 0 0
(x’y’,z’)=(x,y,z) 0 1 0 ----(49)
0 0 -1

In terms of HCS (Homogenous coordinate systems), equation (49) becomes

1 0 0 0
(x’y’,z’,1)=(x,y,z,1) 0 1 0 0
0 0 -1 0
0 0 0 1

That is, P’=P.Mxy -----(50)

Similarly, the mirror reflection about yz plane shown in Figure 12 can be represented
as:

x’=-x
Myz = y’=y -----(51)
z’=z Z

P(−x, y, z)

Y
P (x, y, z)
X Figure 12
In matrix form,

–1 0 0
(x’y’,z’)=(x,y,z) 0 1 0 ------(52)
0 0 1

In terms of HCS, equation (52) becomes

–1 0 0 0
(x’y’,z’,1)=(x,y,z,1) 0 1 0 0
0 0 1 0
0 0 0 1
36
 2-D and 3-D
Transformations
That is, P’=P. Myz ; ------(53)

and similarly, the reflection about xz plane, shown in Figure 13, can be presented as:
x’=x
Mxz = y’= − y -------(54)
z’=z

In matrix form,
Z
1 0 0
(x’y’,z’)=(x,y,z) 0 –1 0 -------(55) P (x, y, z)
0 0 1

In terms of HCS, equation (55) becomes

Y

1 0 0 0
(x’y’,z’,1)=(x,y,z,1) 0 –1 0 0
Figure 13
0 0 1 0 X
0 0 0 1
That is, P’=P. Mxz --------(56)

1.6 SUMMARY

In this unit, the following things have been discussed in detail:

• Various geometric transformations such as translation, rotation, reflection, scaling

and shearing.
• Translation, Rotation and Reflection transformations are used to manipulate the
given object, whereas Scaling and Shearing transformation changes their sizes.
• Translation is the process of changing the position (not the shape/size) of an
object w.r.t. the origin of the coordinate axes.
• In 2-D rotation, an object is rotated by an angle θ. There are two cases of 2-D
rotation: case1- rotation about the origin and case2- rotation about an arbitrary
point. So, in 2-D, a rotation is prescribed by an angle of rotation θ and a centre of
rotation, say P. However, in 3-D rotations, we need to mention the angle of
rotation and the axis of rotation.
• Scaling process is mainly used to change the shape/size of an object. The scale
factors determine whether the scaling is a magnification, s>1, or a reduction, s<1.
• Shearing transformation is a special case of translation. The effect of this
transformation looks like “pushing” a geometric object in a direction that is
parallel to a coordinate plane (3D) or a coordinate axis (2D). How far a direction
is pushed is determined by its shearing factor.
• Reflection is a transformation which generates the mirror image of an object. For
reflection we need to know the reference axis or reference plane depending on
whether the object is 2-D or 3-D.
• Composite transformation involves more than one transformation concatenated
into a single matrix. This process is also called concatenation of matrices. Any
transformation made about an arbitrary point makes use of composite
transformation such as Rotation about an arbitrary point, reflection about an
arbitrary line, etc.
• The use of homogeneous coordinate system to represent the translation
transformation in matrix form, extends our N-coordinate system with (N+1)
coordinate system.

37
 Transformations
• The transformations such as translation, rotation, reflection, scaling and shearing
can be extended to 3D cases.

1.7 SOLUTIONS/ANSWERS
Check Your Progress 1

1) Matrix representation are standard method of implementing transformations in

computer graphics. But unfortunately, we are not able to represent all the
transformations in a (2 x 2) matrix form; such as translation. By using
Homogeneous coordinates system (HCS), we can represent all the transformations
in matrix form. For translation of point (x, y) → (x +tx, y + ty), it is not possible to
represent this transformation in matrix form. But, now in HCS;

 1 0 0
(x ′, y ′, 1) = (x, y,1)  0 1 0
t x t y 1
 
The advantage of introducing the matrix form for translation is that we can now
build a complex transformation by multiplying the basic matrix transformation.
This is an effective procedure as it reduces the computations.

2) The translation factor, tx and ty can be obtained from new old coordinates of vertex
C.
tx = 6 −1 = 5
ty = 7 −1 = 6
The new coordinates [A′ B′ C′ D′] = [A B C D] . Tv

A ′  x 1′ y1′ 1 0 0 1 5 6 1
    1 0 0  5 7
B′  x ′2 y ′2 1 0 1 1  1
=  . 0 1 0 = 
C ′  x ′3 y ′3 1 1 1 1 6 7 1
    5 6 1  
D ′  x ′4 y ′4 1 1 0 1 6 6 1
Thus A′ = (5, 6), B′ = (5, 7), C′ = (6, 7) and D′ = (6, 6)

3) The new coordinate P′ of a point P, after the Rotation of 45o is:

P′ = P.R 45O
 Cos 45 o Sin 45 o 0  1 / 2 1 / 2 0
 
(x ′, y′,1) = (x, y,1) − Sin 45 o 
Cos 45 o 0 = (x , y, 1)  − 1 / 2 1 / 2 0
 0 0 1  0 0 1
  

 1
=  (x − y ), 1
(x + y, 1) = (0, 6 / )
2, 1
 2 2 

Now, this point P′is again translated by tx = 5 and ty = 6. So the final coordinate
P′′ of a given point P, can be obtained as:

1 0 0
(x ′′, y′′, 1) = (x ′, y ′, 1) . 0 1 0
5 6 1

38
 2-D and 3-D
Transformations
1 0 0 
 6 
= ( )
0, 6 / 2,1 . 0 1 0  =  5, + 6, 1
5 6 1  
6 

6
Thus P′′ (x′′, y′′) = (5, + 6)
2
 Cosθ Sinθ   Cos(− θ) Sin (− θ)  Cosθ − Sinθ 
4) R θ =   R −θ =   =  
 − Sinθ Cosθ   − Sin (− θ) Cos(θ)   Sinθ Cosθ 

 Cosθ Sinθ   Cosθ − Sinθ   1 0

Q R θ . R −θ =     =   = Identity matrix
 − Sinθ Cosθ   − Sinθ Cosθ   0 1 

Therefore, we can say that R θ . R −θ are inverse because R θ . R −θ = I. So

R −θ = R θ−1 i.e., inverse of a rotation by θ degree is a rotation in the opposite
direction.

Check Your Progress 2

1) Scaling transformation is mainly used to change the size of an object. The scale
factors determines whether the scaling is a compression, S < 1 or a enlargement,
S > 1, whereas the effect of shearing is “pushing” a geometric object is a direction
parallel to the coordinate axes. Shearing factor determines, how far a direction is
pushed.

a 0  c 0   a.c 0 
2) S a,b =   , Sc,d =   and S ac,bd =  
0 b 0 d  0 b.d 
since
 a 0   c 0   a.c 0 
Sa,b. Sc,d =   .   =   – (1)
0 b 0 d 0 b.d 

c 0  a 0   c.a 0 
and Sc,d. Sa,b =   .   =   – (2)
0 d  0 b  0 d.b 

from (1) and (2) we can say:

Sa,b. Sc,d = Sc,d. Sa,b = Sac, bd

3)
a) Shift an image to the right by 3 units

 0 0 0
 
∴ S =  0 1 0
 3 0 1
 
b) Shift the image up by 2 units and down by 1 units i.e. Sx = Sx + 2 and
Sy = Sy – 1

39
 Transformations
 (Sx + 2) 0 0   1 0 0  1 0 0
     
∴ S = 0 (Sy – 1) 0  ∴S=  0 1 0  0 1 0 =
0 0 1  0 2 1   0 −1 1 
    
 1 0 0
 
 0 1 0
 0 1 1
 

c) Move the image down 2/3 units and left 4 units

1 0 0
 
∴S =  0 1 0
 −4 −2 / 3 1 
 

 Sx 0  cos θ sin θ 
4) SSx, Sy =   and Rθ =  
 − sin θ cos θ 
0 Sy  

we have to find out condition under which SSx, Sy. Rθ = Rθ. Ss, Sy

 Sx o   cos θ sin θ   Sx.cosθ S x .sin θ 

so SSx, Sy. Rθ =  . =
  — (1)
o Sy   − sin θ cos θ   − S y .sin θ Sy.cosθ 
 cos θ sin θ   Sx o   cosθ .Sx Sinθ .S y 
and Rθ. SSx,Sy =  .  =  — (2)
 − sin θ cos θ  o Sy   − S x .Sinθ cosθ .Sy 

In order to satisfy SSx, Sy. Rθ = Rθ. SSx,Sy

We have Sy. sin θ = sinθ.Sx ⇒ either sin θ = 0 or θ = n π, where n is an

integer.
sin θ (Sy – Sx) = 0 or Sx = Sy i.e. scaling transform is uniform.

1 o  1 b  1 b 
5) No, since Shx (a). Shy(b) =   .   =   — (1)
 a 1  o 1  a ab +1
1 b  1 o  1 + ba b 
Shy (b).Shx (a) =     =   — (2)
 o 1   a 1  a 1 
1 b 
and Shxy (a, b) =  
 a 1

from (1) , (2) and (3), we can say that

Shxy (a, b) ≠ Shx (a) . Shy (b) ≠ Shy (b). Shx (a)

Check Your Progress 3

0 1  1 0 
1) M y = x =  , M x =   and
1 0 0 − 1
 Cos 90 o Sin 90 o   0 1
Counter clockwise Rotation of 90o; R 90o =  o
=  
− Sin 90 Cos 90  − 1 0
o

40
 2-D and 3-D
We have to show that Transformations

My = x = Mx . R 90o
1 0   0 1  0 1 
Since Mx . R 90o =   .  =   = My = x
0 − 1 − 1 0 1 0

Hence, a reflection about the line y = x, is equivalent to a reflection relative to the

x-axis followed by a counter clockwise rotation of 90o.

2) The required single (3 x 3) homogeneous transformation matrix can be obtained

as follows:

 2 0 0  1 0 0  2 0 0
a) T = S2,2 . Ttx-1, ty = 0 2 0 .  0 1 0  =  0 2 0 
     
0 0 1  −1 0 1   −1 0 1 

3/ 2 0 0   Cos 90 Sin 90 0 

o o

 
b) T= S 3 . R 90o =  0 1 0  .  − Sin 90o Cos 90o 0 
sx + , sy
2  0 0 1   0 0 1 
 

3/ 2 0 0   0 1 0   0 3/ 2 0
=  0 1 0  .  −1 0 0  =  −1 0 0 
 0 0 1   0 0 1   0 0 1 

 Cos90o Sin90o 0  3/ 2 0 0 

 
c) T = R90o . S 3 =  − Sin90o Cos90o 0  .  0 1 0 
Sx + , S y
2  0 0 1   0 0 1 
 

 0 1 0 3/ 2 0 0 
= − 1 0 0 .  0 1 0 
 0 0 1  0 0 1 

 0 1 0
= 3/ 2 0 0 
 0 0 1 
 
1 0 0   1/ 2 1/ 2 0 
   
T = T −1 1 . R45o =  0 1 0  .  −1/ 2 1/ 2 0
tx , t y +
1  
2 2
1 1  0 0 1
   
 2 2 
 1/ 2 1/ 2 0 
 
 −1/ 2 1/ 2 0 
 
 1 2 0 1


41
 Transformations
3) Let OP be given line L, which makes an angle θ with respect to

P
L

y = ax
θ
O

The transformation matrix for reflection about an arbitrary line y = mx + c is

(see equation 25).

1 − m2 2m 
 2 0 
 m +1 m2 + 1 
 2m m2 − 1 
ML =  2 0  where m = tan θ
m +1 m2 + 1 
 − 2 cm − 2C 
 2 1 
 m + 1 m2 + 1 

For line y = ax; m = tanθ = a and intercept on y-axis is 0 i.e. c = 0. Thus,

transformation matrix for reflection about a line y = ax is:

1 − a 2 2a 
 2 
 a + 1 a 2 + 1 0
2

 2.a a − 1 
M L = M y =ax =  2 0 where a = tan θ = m
a +1 a 2 +1 
 1
 0 0 
 
4) The equation of the line passing through the points (1,3) and (−1, −1) is obtained
as:
y = 2x + 1 (1)
y
P2 (1, 3)
L
L’
(0, 1)
θ C (3, 2)
x
B (2, 0)

P1 (−1, 1)

Figure (a)

If θ is the angle made by the line (1) with the positive x-axis, then
1 2
tanθ = 2 ⇒ Cosθ = and Sinθ
2 5
To obtain the reflection about the line (1), the following sequence of transformations
can be performed:

1) Translate the intersection point (0, 1) to the origin, this shift the line L to L′
2) Rotate the shifted line L′ by −θo (i.e. clockwise), so that the L′aligns with the x-
axis.

42
 2-D and 3-D
3) Perform the reflection about x-axis. Transformations
4) Apply the inverse of the transformation of step (2).
5) Apply the inverse of the transformation of step (1).

By performing step 1 – step 5, we get

M L = TV . R θ M X . R θ−1 . TV−1

1 0 0 1 / 5 − 2 / 5 0 1 0 0  1 / 5 2 / 5 0 1 0 0
   
= 0 1 0 . 2 / 5 1 / 5 0 . 0 − 1 0 . − 2 / 5 1 / 5 0 . 0 1 0
0 − 1 1  0 0 1 0 0 1  0 0 1 0 1 1
   

 − 3 / 5 4 / 5 0
 
=  4 / 5 3 / 5 0
 − 4 / 5 2 / 5 1
 

So the new coordinates A′B′C′ of the reflected triangle ABC can be found as:
[A′ B′ C′] = [ABC] . ML

0 3 1 − 3 / 5 4 / 5 0 8 / 5 11 / 5 1
 
 
= 2 0 1 .  4 / 5 3 / 5 0 =  − 2 2 1
3 2 1  − 4 / 5 2 / 5 1  − 1 4 1
 

 11 
Thus, A′ =  8 / 5,  , B′ = (− 2, 2) and C′ = (−1, 4), which is shown in Figure (b).
 5

Y
C’ 4
3

B’ 2 A’
1
X

–4 –3 –2 –1 1 2 3 4

Figure (b)

43
Transformations
UNIT 2 VIEWING TRANSFORMATION
Structure Page Nos.

2.0 Introduction 44
2.1 Objectives 45
2.2 Projections 45
2.2.1 Parallel Projection 47
2.2.1.1 Orthographic and Oblique Projections 51
2.2.1.2 Isometric Projections 61
2.2.2 Perspective Projections 65
2.3 Summary 83
2.4 Solutions/Answers 84

2.0 INTRODUCTION

In unit 1, we have discussed the geometric transformations such as Translation,

Rotation, Reflection, Scaling and Shearing. Translation, Rotation and Reflection
transformations are used to manipulate the given object, whereas Scaling and
Shearing transformations are used to modify the shape of an object, either in 2-D or in
3-Dimensional.

A transformation which maps 3-D objects onto 2-D screen, we are going to call it
Projections. We have two types of Projections namely, Perspective projection and
Parallel projection. This categorisation is based on the fact whether rays coming from
the object converge at the centre of projection or not. If, the rays coming from the
object converge at the centre of projection, then this projection is known as
Perspective projection, otherwise it is Parallel projection. In the case of parallel
projection the rays from an object converge at infinity, unlike perspective projection
where the rays from an object converge at a finite distance (called COP).

Parallel projection is further categorised into Orthographic and Oblique projection.

Parallel projection can be categorized according to the angle that the direction of
projection makes with the projection plane If the direction of projection of rays is
perpendicular to the projection plane then this parallel projection is known as
Orthographic projection and if the direction of projection of rays is not perpendicular
to the projection plane then this parallel projection is known as Oblique projection.
The orthographic (perpendicular) projection shows only the front face of the given
object, which includes only two dimensions: length and width. The oblique projection,
on the other hand, shows the front surface and the top surface, which includes three
dimensions: length, width, and height. Therefore, an oblique projection is one way to
show all three dimensions of an object in a single view.

Isometric projection is the most frequently used type of axonometric projection,

which is a method used to show an object in all three dimensions (length, width, and
height) in a single view. Axonometric projection is a form of orthographic projection
in which the projectors are always perpendicular to the plane of projection.

44
 Viewing
Transformations
2.1 OBJECTIVES

After going through this unit, you should be able to:

• define the projection;
• categorize various types of Perspective and Parallel projections;
• develop the general transformation matrix for parallel projection;
• describe and develop the transformation for Orthographic and oblique parallel
projections;
• develop the transformations for multiview (front, right, top, rear, left and bottom
view) projections;
• define the foreshortening factor and categorize the oblique projection on the basis
of foreshortening factors;
• derive the transformations for general perspective projection;
• describe and derive the projection matrix for single-point, two-point and three-
point perspective transformations, and
• identify the vanishing points.

2.2 PROJECTIONS
.
Given a 3-D object in a space, Projection can be defined as a mapping of 3-D object
onto 2-D viewing screen. Here, 2-D screen is known as Plane of projection or view
plane, which constitutes the display surface. The mapping is determined by projection
rays called the projectors. Geometric projections of objects are formed by the
intersection of lines (called projectors) with a plane called plane of projection /view
plane. Projectors are lines from an arbitrary point, called the centre of projection
(COP), through each point in an object. Figure 1 shows a mapping of point P(x,y,z)
onto its image P′(x’,y’,z’) in the view plane.

y
P(x, y, z)

Projector
P’(x’,y’,z’)

x
z
Figure 1

If, the COP (Center of projection) is located at finite point in the three-space, the
result is a perspective projection. If the COP is located at infinity, all the projectors are
parallel and the result is a parallel projection. Figure 2(a)-(b) shows the difference
between parallel and perspective projections. In Figure 2(a), ABCD is projected to
A’B’C’D’ on the plane of projection and O is a COP. In the case of parallel
projection the rays from an object converges at infinity, the rays from the object
become parallel and will have a direction called ”direction of projection”.

45
Transformations

Figure 2(a): Perspective projection

Figure 2(b): Parallel projection

Taxonomy of Projection

There are various types of projections according to the view that is desired. The
following Figure 3 shows taxonomy of the families of Perspective and Parallel
Projections. This categorisation is based on whether the rays from the object converge
at COP or not and whether the rays intersect the projection plane perpendicularly or
not. The former condition separates the perspective projection from the parallel
projection and the latter condition separates the Orthographic projection from the
Oblique projection.

46
 Viewing
Projection Transformations

Perspective (Rays from the Parallel (Rays from the

object converges at the COP) object are parallel)

Single-point Two-point Three-point Orthographic (parallel rays intersect

view plane perpendicularly)
Oblique (parallel rays intersect
view plane not perpendicularly)

Multiview (view plane Axonometric (vie plane not Cavalier Cabinet

parallel to principal axes) parallel to principal axes)

Top Bottom Isometric Diametric Trimetric

Front Rear Right-Side Left-Side

Figure 3: Taxonomy of projection

The direction of rays is very important only in the case of Parallel projection. On the
other hand, for Perspective projection, the rays converging at the COP, they do not
have a fixed direction i.e., each ray intersects the projection plane with a different
angle. For Perspective projection the direction of viewing is important as this only
determines the occurrence of a vanishing point.

2.2.1 Parallel Projection

Parallel projection methods are used by engineers to create working drawings of an

object which preserves its true shape. In the case of parallel projection the rays from
an object converge at infinity, unlike the perspective projection where the rays from
an object converse at a finite distance (called COP).

If the distance of COP from the projection plane is infinite then parallel projection (all
rays parallel) occurs i.e., when the distance of COP from the projection plane is
infinity, then all rays from the object become parallel and will have a direction called
“direction of projection”. It is denoted by d=(d1,d2,d3), which means d makes
unequal/equal angle with the positive side of the x,y,z axes.

Parallel projection can be categorised according to the angle that the direction of
projection makes with the projection plane. For example, in Isometric projection, the
direction of projection d=(d1,d2,d3) makes equal angle (say α) with all the three-
principal axes (see Figure 4).
47
Transformations

View plane
y
d = (d1, d2, d3)

α
β
z x
χ

Figure 4: Direction of projection

Rays from the object intersect the plane before passing through COP. In parallel
projection, image points are found as the intersection of view plane with a projector
(rays) drawn from the object point and having a fixed direction.(see Figure 5).

Direction of projection
y
d
P1 (x,y,z)

P’,(x’,y’,z’)

P2
P2’

Figure 5: Parallel projection

Parallel rays from the object may be perpendicular or may not be perpendicular to the
projection plane. If the direction of projection d=(d1,d2,d3) of the rays is
perpendicular to the projection plane (or d has the same direction as the view plane
normal N), we have Orthographic projection otherwise Oblique projection.

Orthographic projection is further divided into Multiview projection and axonometric

projection, depending on whether the direction of projection of rays is parallel to any
of the principal axes or not. If the direction of projection is parallel to any of the
principal axes then this produces the front, top and side views of a given object, also
referred to as multiview drawing (see Figure 8).

Axonometric projections are orthographic projection in which the direction of

projection is not parallel to any of the 3 principle axes. Oblique projections are non-
orthographic parallel projections i.e., if the direction of projection d=(d1,d2,d3) is not
perpendicular to the projection plane then the parallel projection is called an Oblique
projection.

48
 Viewing
Transformation for parallel projection Transformations

Parallel projections (also known as Orthographic projection), are projections onto one
of the coordinate planes x = 0, y = 0 or z = 0. The standard transformation for parallel
(orthographic) projection onto the xy-plane (i.e. z=0 plane) is:
Ppar,z =
x ' = x

y ' = y
z ' = 0

In matrix form:
1 0 0 0
 
0 1 0 0
Ppar,z =  ------------------------------(1)
0 0 0 0
 
0 0 0 1

Thus, if P(x,y,z) be any object point in space, then projected point P’(x’y’z’) can be
obtained as:
1 0 0 0
 
0 1 0 0
(x’,y’,z, 1) = (x, y, z, 1)  ------------------------------(2)
0 0 0 0
 
0 0 0 1 

P’h =Ph.Ppar,z ------------------------------(3)
Example1: Derive the general transformation of parallel projection onto the xy-plane
in the direction of projection d=aI+bJ+cK.
Solution: The general transformation of parallel projection onto the xy-plane in the
direction of projection d=aI+bJ+cK, is derived as follows(see Figure a):

Let P(x,y,z) be an object point, projected to P’(x’,y’,z’) onto the z’=0 plane.
From Figure (a) we see that the vectors d and PP’ have the same direction.
This means that

PP’=k.d , comparing components, we have:

x’-x=k.a
y’-y=k.b
z’-z=k.c
Since z’=0 on the projection plane, we get k=-z/c.
Thus,
x’=x-a.z/c
y’=y-b.z/c
z’=0 z V= aI + bj +ck
P (x, y, z)

0
y
x
P’ (x’, y’, 0)
Figure (a)
49
Transformations In terms of homogeneous coordinates, this equation can be written as:

 1 0 0 0
 
 0 1 0 0
(x’,y’,z’1) = (x, y, z, 1)  ------------------------------(4)
− a/c −b/c 0 0
 
 0 0 0 1 


That is, P’h =Ph.Ppar,z , where Ppar,z is the parallel projection with the direction of
projection d along the unit vector k.

Example 2: Derive the general transformation for parallel projection onto a given
view plane, where the direction of projection d=aI+bJ+cK is along the normal
N=n1I+n2J+n3K with the reference point R0(x0,y0,z0).

Solution: The general transformation for parallel projection onto the xy-plane in the
direction of projection Figure (b)
v = a I + bJ + ck, denoted by P par, V, N, Ro, consists of the following steps:

1) Translate the view reference point Ro of the view plane to the origin, by T-Ro
2) Perform an alignment transformation An so that the view normal vector N of the
view points in the direction K of the normal to the xy-plane. The direction of
projection vector V is transformed to new vector V’ = AnV.
3) Project onto the xy-plane using P par, v’
4) Align k back to N, using An.
5) Translate the origin back to Ro, by TRo

z
N = n1I + n2J + n3k
k View plane

V = aI +bJ +ck
P’(x’,y’,z’)

R0 ’(x0, y0, z0)

o
y
P(x,y,z)
Figure (b)
x

So
Ppar, V, N, Ro = T–Ro AN–1 . Ppar, v’. An. TRo

 λ − n 1n 2 − n1n 3 
 0
 1 0 0 0 N N 
   n3 n2 
 0 1 0 0  0 0
=  λ λ
0 0 1 0  
   n1 n2 n3

 − x0 − y0 − z0 1 0
N N N 
 
 0 0 0 1

50
 Viewing
 λ n2  Transformations
1 0 0 0  |N| 0 0 
   |N| 
−a  1 0 0 0
1 0 0  − n1n2 n3 n2   
 c   0   0 1 0 0
−a   |N| λ |N| 
−b  0 0 1 0
 0 0  − n1n3 n2 n3   
 c c   0  x0 y0 z0 1 
 0 0 0 1   λ | N| |N| |N| 
 
 0 0 0 1
where λ =

n22 + n32 and λ ≠ 0.

After multiplying all the matrices, we have:

 d1 − an1 − bn1 − cn1 0

 
 − an2 d1 − bn 2 − cn2 0
P par, V, N, R0 =  ---------------------------(5)
− an3 − bn3 d1 − cn3 0
 
 ad 0 bd 0 c.do d1 


Where d0 = n1 x0 + n2 y0 + n3 z0 and
d1 = n1a + n2b + n3c

Note: Alignment transformation, An, refer any book for computer graphic.

2.2.1.1 Orthographic and Oblique Projections

Orthographic projection is the simplest form of parallel projection, which is

commonly used for engineering drawings. They actually show the ‘true’ size and
shape of a single plane face of a given object.

If the direction of projection d=(d1,d2,d3)has the direction of view plane normal N (or
d is perpendicular to view plane), the projection is said to be orthographic. Otherwise
it is called Oblique projection. The Figure 6 shows the orthographic and oblique
projection.

We can see that the orthographic (perpendicular) projection shows only front surface
of an object, which includes only two dimensions: length and width. The oblique
projection, on the other hand, shows the front surface and the top surface, which
includes three dimensions: length, width, and height. Therefore, an oblique
projection is one way to show all three dimensions of an object in a single view

Figure 6: Orthographic and oblique projection

51
Transformations Orthographic projections are projections onto one of the coordinate planes x=0, y=0or
z=0. The matrix for orthographic projection onto the z=0 plane (i.e. xy-plane) is:

1 0 0 0
 
0 1 0 0
Ppar,z =  ------------------------------(6)
0 0 0 0
 
0 0 0 1 


Note that the z-column (third column) in this matrix is all zeros. That is for
orthographic projection onto the z=0 plane, the z-coordinates of a position vector is
set to zero. Similarly, we can also obtain the matrices for orthographic projection onto
the x=0 and y=0 planes as:

0 0 0 0
 
0 1 0 0
Ppar,x =  ------------------------------(7)
0 0 1 0
 
0 0 0 1 

and
1 0 0 0
 
0 0 0 0
Ppar,y =  ------------------------------(8)
0 0 1 0
 
0 0 0 1 

For example, consider the object given in Figure 6(a). The orthographic projections of
this object onto the x=0, y=0 and z=0 planes from COP at infinity on the +x-, +y- and
+z-axes are shown in Figure 7 (b)-(d).

x
z
z Figure 7(a) Figure 7(b)

y y

Figure 7(c) x z Figure 7(d)

52
 Viewing
Transformations
A single orthographic projection does not provide sufficient information to visually
and practically reconstruct the shape of an object. Thus multiple orthographic
projections are needed (known as multiview drawing). In all, we have 6 views:
1) Front view
2) Right-side view
3) Top-view
4) Rear view
5) Left-side view
6) Bottom view
The Figure 8 shows all 6 views of a given object.

Figure 8: Multiview orthographic projection

The front, right-side and top views are obtained by projection onto the z=0, x=0 and
y=0 planes from COP at infinity on the +z-, +x-, and +y-axes.
The rear, left-side and bottom view projections are obtained by projection onto the
z=0, x=0, y=0 planes from COP at infinity on the –z-, -x and –y-axes(see Figure 8).
All six views are normally not required to convey the shape of an object. The front,
top and right-side views are most frequently used.
The direction of projection of rays is shown by arrows in Figure 9.
+y

Top –z

Left-side rear

–x x
right-side

z front
Figure 9: Direction of projection of rays in multiview drawing
–y 53
Transformations The projection matrices for the front, the right-side and top views are given by:
Pfront = Ppar,z=diag(1,1,0,1)
Pright = Ppar,x=diag(0,1,1,1)
Ptop = Ppar,y=diag(1,0,1,1)
It is important to note that the other remaining views can be obtained by combinations
of reflection, rotation and translation followed by projection onto the z=0 plane from
the COP at infinity on the +z-axis. For example: the rear view is obtained by
reflection through the z=0 plane, followed by projection onto the z=0 plane.
Prear=Mxy. Ppar,z

1 0 0 0 1 0 0 0 1 0 0 0
     
0 1 0 0 0 1 0 0 0 1 0 0
= =  -------------------------(9)
0 0 −1 0 0 0 0 0  0 0 0 0
     
0 0 0 1  0 0 0 1  0 0 0 1 
  
Similarly, the left-side view is obtained by rotation about the y-axis by +900, followed
by projection onto the z=0 plane.
Pleft=[Ry] 900.Ppar,z

 cos 90 0 − sin 90 0 1 0 0 0 0 0 0 0
     
 0 1 0 0 0 1 0 0 0 1 0 0
= = ------------------(10)
sin 90 0 cos 90 0 0 0 0 0 1 0 0 0
     
 0 1  0 0 1 
 0 0  0 0 1   0 0

And the bottom view is obtained by rotation about the x-axis by -900, followed by
projection onto the z=0 plane.
Pbottom=[Rx] 900.Ppar,z

1 0 0 0  1 0 0 0  1 0 0 0
    
 0 cos(−90) sin( −90) 0  0 1 0 0  0 0 0 0
= = -----------(11)
0 − sin( −90) cos(−90) 0  0 0 0 0  0 1 0 0
    
0 1   0 0
 0 0 0 1   0 0 0 1 

Example 3: Show all the six views of a given object shown in following Figure. The
vertices of the object are A(4,0,0), B(4,4,0), C(4,4,8), D(4, 0, 4), E (0,0,0), F(0,4,0),
G(0,4,8), H(0,0,4).
Solution: We can represent the given object in terms of Homogeneous-coordinates of
its vertices as:
A  4 0 0 1
 
B  4 4 0 1 F B
C  4 4 8 1 C
 
D  4 0 4 1 G
V= [ABCDEFGH] =  
E  0 0 0 1
F  0 4 0 1 A x
E
 
G  0 4 8 1
H  0 0 4 1
H D
Figure (c)
z
54
 Viewing
Transformations
(1) If we are viewing from the front, then the new coordinate of a given object can be
found as:

P’n,z = Pn. Pfront

A'  x '1 y '1 1 A'  4 0 0 1

  4 0 1 0  
B'  x '2 y '2 1   B'  4 4 0 1
4 4 1 0
C'  x '3 y '3 1 4 1  1 0 0 0  C'  4 4 0 1
  
4

8   
D'  x '4 y '4 1 4 0 1  0
4 1 0 0  D'  4 0 0 1
= . =
E'  1  0 0  E'  0 1 

x '5 y '5 0 0 1  0
0 0 0 0
 
F'  x '6 y '6 1
0 4 8 1  0 0 0 1  F'  0 4 0 1
  
  0 4 8 1  
G'  x '7 y '7 1 0 G'  0 4 0 1
 0 4 1
H'  x '8 y '8 1  H'  0 0 0 


from matrix, we can see that

A′ = D′, B′ = C′, E′ = H′, F′ = G′, Thus we can see only C′D′G′H′
as shown in Figure d y

G′ C′

H′ D′
x

Figure d

(2) If we are viewing from right-side, then

A 4 0 0 1 A'  0 0 0 1
   
B 4 4 0 1 B'  0 4 0 1
C 4 4 8 1 0 0 0 0  C'  0 4 8 1
     
D 4 0 4 1 0 1 0 0 D 0 0 4 1
P’n,x = V. Pright = 0 . =
E  0 0 1 0 0 1 0  E'  0 0 0 1
 
F 0 4 0 1 0 0 0 1  F'  0 4 0 1

   
G 0 4 0 1 G'  0 4 8 1
H 0 0 4 1 H'  0 0 4 1


Here, we can see that A’ =E’, B’ = F’, C’ = G’ and D’ = H’.

Thus, we can see only A’B’C’D’ as shown in Figure e.
y

B′ C′

z
A′ Figure e D′
55
Transformations
(3) if we are viewing from top, then
A4 0 0 1 A' 4 0 0 1
B4 4 0 1 B' 4 0 0 1

C4 4 8 1  1 0 0 0  C' 4 0 8 1
     
D4 0 4 1  0 0 0 0  D' 4 0 4 1
P’n,y = Pn. Ptop = . =
E0 0 0 1  0 0 1 0  E' 0 0 0 1
     
F0 4 0 1  0 0 0 1  F' 0 0 0 1
G 0 4 8 1 G' 0 0 8 1
   
H 0 0 4 1 H' 0 0 4 1

Here, we can see that A’ = B’, E’ = F’, C’ ≠ D’ and G’ ≠ H’

Thus we can see only the square B’F’G’C’ but the line H’D’ is hidden an shown in
Figure f. z

G′ C′

H′ D′

x
F′ B′
Figure f

Similarly we can also find out the other side views like, rear left-side and bottom
using equation – 1, 2, 3

 Check Your Progress 1

1) Define the following terms related with Projections with a suitable diagram:

a) Center of Projection (COP)

b) Plane of projection/ view plane
c) Projector
d) Direction of projection
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
2) Categories the various types of parallel and perspective projection.
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………

56
 Viewing
3) In orthographic projection Transformations
a) Rays intersect the projection plane.
b) The parallel rays intersect the view plane not perpendicularly.
c) The parallel rays intersect the view plane perpendicularly.
d) none of these
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………

Oblique projection

If the direction of projection d=(d1,d2,d3) of the rays is not perpendicular to the view
plane(or d does not have the same direction as the view plane normal N), then the
parallel projection is called an Oblique projection (see Figure 10).
y Direction of projection
Direction of projection
d
view plane
d = (d1 , d 2 , d 3 ) A
A′

B′
view plane
B
z N
x

Figure 10 (a): Oblique projection Figure 10 (b): Oblique projection

In oblique projection only the faces of the object parallel to the view plane are shown
at their true size and shape, angles and lengths are preserved for these faces only.
Faces not parallel to the view plane are discarded.

In Oblique projection the line perpendicular to the projection plane are foreshortened
(shorter in length of actual lines) by the direction of projection of rays. The direction
of projection of rays determines the amount of foreshortening. The change in length of
the projected line (due to the direction of projection of rays) is measured in terms of
foreshortening factor, f.

Foreshortening factors w.r.t. a given direction

Let AB and CD are two given line segments and direction of projection d=(d1,d2,d3).
Also assumed that AB║CD║d . Under parallel projection, let AB and CD be
projected to A’B’ and C’D’, respectively.

Observation:
i) A’B’║C’D’ will be true, i.e. Parallel lines are projected to parallel lines, under
parallel projection.
ii) |A’B’| ⁄ |AB| ═ |C’D’| ⁄ |CD| must be true, under parallel projection.

This ratio (projected length of a line to its true length) is called the foreshortening
factor w.r.t. a given direction.

57
Transformations Mathematical description of an Oblique projection (onto xy-plane)

In order to develop the transformation for the oblique projection, consider the
Figure 10. This figure shows an oblique projection of the point A (0, 0, 1) to position
A’(x’,y’,0) on the view plane (z=0 plane). The direction of projection d=(d1,d2,d3).

Oblique projections (to xy-plane) can be specified by a number f and an angle θ. The
number f, known as foreshortening factor, indicates the ratio of projected length
OA’of a line to its true length. Any line L perpendicular to the xy-plane will be
foreshortened after projection.

θ is the angle which the projected line OA’(of a given line L perpendicular to xy-
plane) makes with the positive x-axis.

The line OA is projected to OA’. The length of the projected line from the origin
=|OA’|

A (0, 0, 1)

d = (d1, d2, d3)

o y'
y
x' β
θ

B
A’ = (x’, y’, ‘0)
x
Figure 11: Oblique projection

Thus, foreshortening factor, f=|OA’|/|OA|=|OA’|, in the z-direction

From the triangle OAP’, we have,

OB=x’=f.cosθ
BA’=y’=f.sinθ

When f = 1, then obligue projection is known as Cavalier projection

Given θ = 45°, then we have

 1 0 0 0
 
 0 1 0 0
Pcav = 
1/ 2 1/ 2 0 0
 
 0 0 0 1 


When f = ½ then obligue projection is called a cabinet projection.

Here θ = 30° (Given), we have

58
 Viewing
Transformations
 1 0 0
0
 
0 1
0 0
Pcab = 
 3 / 4 1/ 4 0 0 
 
 0 0 0 1 

we can represent a given unit cube in terms of Homogeneous coordinates of the

A 0 0 0 1
B 1 0 0 1

C 1 1 0 1
 
D 0 1 0 1
vertices as:V = [A B C D E F G H] =
E 0 1 1 1
 
F 0 0 1 1
G 1 0 1 1
 
H 1 1 1 1 

i) To draw the cavalier projection, we find the image coordinates of a given unit cube
as follows:
A'  0 0 0 1
A 0 1  0 1
0 0
B'  1 0
B 1 0 0 1 
  C'  1 1 0 1

C 1 1 0 1  1 0 0 0
 0 1 0 1
D 0 1 0 1  0 1 0 0 D'  
P’ = V. Pcav =  .   = 2
E 0 1 1 1  1 / 2 1 / 2 0 0  E'  2 / 2 (1 + 0 1
F 0 0 1 1   0 0 0 1 
 2 
  F'  2 / 2 2/2 0 1
G 1 0 1 1
G' (1 + 2 / 2 2/2 0 1

H 1 1 1 1   
H' (1 + 2 / 2 (1 + 2 / 2 0 1

Hence, the image coordinate are:

A’ = (0, 0, 0), B’ = (1, 0, 0), C’ = (1, 1, 0), D’ = (0, 1, 0) E’ = (√2/2 , 1 + √2/2, 0)

F’ = (√2/2, √2/2, 0), G’ = ( 1 + √2/2, √2/2, 0), H’ = (1 + √2/2, 1 + √2/2, 0)

Thus, cavalier projection of a unit cube is shown in Figure 11(a).

y
2
E′ H′

1 C′
D′

G′
F′
B′
1 Figure 11(a)
A′ x
2

59
Transformations To determine projection matrix for oblique projection, we need to find the direction
vector d. Since vector PP’and vector d have the same direction. Thus, PP’=d

Thus, x’– 0=d1= f.cosθ

y’– 0=d2= f.sinθ
z’–1=d3

As z’=0 on the xy-plane, d3 = –1,

Since, Oblique projection is a special case of parallel projection, thus, we can

transform the general transformation of parallel projection for Oblique projection as
follows:

 1 0 0 0  1 0 0 0
   
 0 1 0 0  0 1 0 0
POblique =  = -----------------(12)
−d /d − d2 / d3 0 0   f . cos θ f . sin θ 0 0
 1 3   
 0 0 0 0   0 0 0 1 


Where, f=foreshortening factor, i.e., the projected length of the z-axis unit vector.
If β is the angle

Between the Oblique projectors and the plane of projection then,

1/f=tan (β) , i.e., f= cot(β) ------------- (13)

θ=angle between the projected line with the positive x-axis.

Special cases:

1) If f=0, then cot (β)=0 that is β=900 , then we have an Orthographic projection.
2) If f=1, the edge perpendicular to projection plane are not foreshortened, then
β=cot-1 (1)=450 and this Oblique projection is called Cavalier projection.
3) If f=1/2 (the foreshortening is half of unit vector), then β=cot-1 (1/2)=63.4350 and
this Oblique projection is called Cabinet projection.

Note: The common values of θ are 30° and 45°. the values of (180° – θ) is also
acceptable.

The Figure 12 shows an Oblique projections for foreshortening factor

f=1,7/8,3/4,5/8,1/2, with θ=450

Figure 12: Oblique projections for f=1,7/8,3/4,5/8,1/2, with θ=450 (from left to right)

Example4: Find the transformation matrix for a) cavalier projection with θ=450 , and
b) cabinet projection with θ=300 c) Draw the projection of unit cube for
each transformation.

Solution: We know that cavalier and cabinet projections are a special case of an
oblique projection. The transformation matrix for oblique projection is:

60
 Viewing
Transformations
 1 0 0 0
 
 0 1 0 0
Pobligue = 
f .cosθ f .sinθ 0 1
 
 0 0 0 1 


(ii) To draw the cabinet projection, we find the image coordinates of a unit cube as:

A'  0 0 0 1
B'  1 0 0 1
 1 1 0 1

C'  
D'  0 1 0 1
P’ V. Pcab =  
E'
 3/4 5/ 4 0 1
F'  3/4 1/ 4 0 1
G' (1 + 3 / 4) 1/ 4 0 0
H' (1 + 3 / 4) 
 5/ 4 0 1

Hence, the image coordinates are:

A’ (0, 0, 0), B’ = (1,0,0), C’ = (1, 1, 0), D’ = (0, 1, 0), E’ = (√3/4, 5/4, 0)
F’ = (√3/4, 1/4, 0), G’ = (1 + √3/4, 1/4, 0), H’ = (1 + √3/4, 5/4, 0)

The following Figure (g) shows a cabinet projection of a unit cube.

2 y
E′ H′

1 C′
D′

F′
G′

B′
x
A′ 1 2
Figure (g)

2.2.1.2 Isometric Projection

There are 3 common sub categories of Orthographic (axonometric) projections:

1) Isometric: The direction of projection makes equal angles with all the three
principal axes.
2) Dimetric: The direction of projection makes equal angles with exactly two of the
three principal axes.
3) Trimetric: The direction of projection makes unequal angles with all the three
principal axes.

Isometric projection is the most frequently used type of axonometric projection, which
is a method used to show an object in all three dimensions in a single view.
Axonometric projection is a form of orthographic projection in which the projectors
are always perpendicular to the plane of projection. However, the object itself,
rather than the projectors, are at an angle to the plane of projection.
61
Transformations Figure13 shows a cube projected by isometric projection. The cube is angled so that
all of its surfaces make the same angle with the plane of projection. As a result, the
length of each of the edges shown in the projection is somewhat shorter than the
actual length of the edge on the object itself. This reduction is called foreshortening.
Since, all of the surfaces make the angle with the plane of projection, the edges
foreshorten in the same ratio. Therefore, one scale can be used for the entire layout;
hence, the term isometric which literally means the same scale.

Construction of an Isometric Projection

In isometric projection, the direction of projection d = (d1,d2,d3) makes an equal angles

with all the three principal axes. Let the direction of projection d = (d1,d2,d3) make
equal angles (say α) with the positive side of the x,y, and z axes(see Figure 13).

Then
i.d=d1=|i|.|d|.cosα => cosα=d1/|d| Z
d=(d1,d2,d3)
similarly
d2=j.d=|j|.|d|.cosα => cosα=d2/|d|
d3=k.d=|k|.|d|.cosα => cosα=d3/|d|
α
so cosα=d1/|d| = d2/|d| = d3/|d| α α
Ö d1= d2 = d3 is true
we choose d1=d2=d3=1 X

Y Figure 13
Thus, we have d =(1, 1, 1)

Since, the projection, we are looking for is an isometric projection => orthographic
projection, i.e, the plane of projection, should be perpendicular to d, so d = n = (1,1,1).
Also, we assume that the plane of projection is passing through the origin.

Ö We know that the equation of a plane passing through reference point

R(x0,y0,z0) and having a normal N = (n1,n2,n3) is: (x – x0).n1 + (y – y0).n2 +
(z –z0).n3=0 -------------------------(14)

Since (n1,n2,n3)=(1,1,1) and

(x0,y0,z0)=(0,0,0)
From equation (14), we have x + y + z = 0
Thus, we have the equation of the plane: x + y + z = 0 and d = (1,1,1)

Transformation for Isometric projection

Let P(x,y,z) be any point in a space. Suppose a given point P(x,y,z) is projected to
P’(x’y’,z’) onto the projection plane x + y + z = 0. We are interested to find out the
projection point P’(x’,y’,z’).

The parametric equation of a line passing through point P(x, y, z) and in the direction
of d (1, 1, 1) is:

P + t.d = (x, y, z) + t. (1,1,1) = (x + t, y + t, z + t) is any point on the line, where

– ∞< t < ∞. The point P’ can be obtained, when t = t*.
Thus P’=(x’,y’,z’)=(x + t*,y + t*,z + t*), since P’ lies on x + y + z = 0 plane.

Ö (x + t*)+(y + t*) + (z + t*)=0

Ö 3.t*=-(x + y + z)
Ö t*=-(x + y + z)/3 should be true.
Ö x’= (2.x – y – z)/3 , y’=(–x +2.y – z)/3 , z’=(– x – y +2.z)/3

62
 Viewing
Transformations
Thus, P’=(x’,y’,z’)=[(2.x –y–z)/3, (–x +2.y– z)/3, (-x-y+2.z)/3] --------------------(15)
In terms of homogeneous coordinates, we obtain
 2 3 − 1 3 1 3 0
 
 − 1 3 2 3 − 1 3 0
(x’, y’, z, 1) = (x, y, z, 1) 
− 1 3 − 1 3 2 3 0
 
 0 0 0 1 


Note: We can also verify this Isometric transformation matrix by checking all the
foreshortening factors, i.e., to check whether all the foreshortening factors (fx, fy, fz)
are equal or not. Consider the points A,B and C on the coordinate axes (see Figure14).
y
B (0, 1, 0)

(0, 1, 0)
x
z (0, 0, 1) A
Figure 14
C
i) Take OA, where O=(0,0,0) and A (1,0,0). Suppose O is projected to O’ and A is
projected to A’
Thus, by using equation (15), we have O’=(0,0,0) and A’=(2/3,-1/3,-1/3).
So |O’A’| = (2 / 3) 2 +(-1/3)2 +(-1/3)2 = 2 / 3 = fx ------------(16)

ii) Take OB, where O = (0,0,0) and B (0,1,0). Suppose O is projected to O’ and B is
projected to B’. Thus by using equation (15), we have O’=(0,0,0) and
B’= (-1/3,2/3,-1/3).
So |O’B’| = − (1 / 3) 2 +(2/3)2 +(-1/3)2 = 2 / 3 = fy -------------(17)

iii) Take OC, where O=(0,0,0) and C(0,0,1). Suppose O is projected to O’ and C is
projected to C’
Thus, by using equation(15), we have O’=(0,0,0) and C’=(-1/3,-1/3,2/3).
So |O’C’|=√(-1/3)2 +(-1/3)2 +(2/3)2 = √2/3=fz --------------(18)
Thus, we have fx=fy=fz, which is true for Isometric projection.
Example 5: Obtain the isometric view of a cuboid, shown in figure. The size of
cuboid is 10x8x6, which is lying at the origin.
z
Solution: The given cuboids can be represented
in terms of Homogeneous coordinates
of vertices as shown in Figure (h): H G

6
D
C

F
8
y
10 E
Figure (h) 63
A B
x
Transformations

A 10 0 0 1
 
B 10 8 0 1
C 10 8 6 1
 
D 10 8 6 1
V = [A B C D E F G H] =  
E 0 0 0 1
F 0 8 0 1
 
G 0 8 6 1
H  0 0 6 1

To draw an Isometric projection, we find the image coordinate of a given cuboid as

follows:
10 0 0 1
 
10 8 0 1
10 8 6 1  2 − 1 − 1 0
 
10 0 6 1 − 1 2 − 1 0
P’ = V. PISO =  . =
 0 0 0 1  − 1 − 1 2 0
0 8 0 1   0 0 0 3
 
0 8 6 1 
0 0 6 1 
 
A'  20 − 10 − 10 3  6.66 − 3.33 − 3.33 1

B'  12 8 − 18 3    2.66 − 6.0 1 
 4 .0
C'  6 0 −6 3  2 0 − 2.0 1
   
D'  14 − 16 2 3 4.66 − 5.33 0.66` 1
= 
E'  0 0 0 3  0 0 0 1
   
F'  − 8 16 − 8 3 − 2.66 5.33 1.33 1
G' − 14 10 4 3 − 4.66 3.33 1.33 1 
   
H'  − 6 − 6 12 3   − 2.0 − 2.0 4.0 1 

Thus, by using this matrix, we can draw an isometric view of a given cuboids.

 Check Your Progress 2

1) When all the foreshortening factors are different, we have

a) Isometric b) Diametric c) Trimetric Projection d) All of these.
……………………………………………………………………………………
……………………………………………………………………………………
……………………………………………………………………………………
……………………………………………………………………………………
2) Distinguish between Orthographic and Oblique parallel projection.
……………………………………………………………………………………
……………………………………………………………………………………
……………………………………………………………………………………
……………………………………………………………………………………

64
 Viewing
3) What do you mean by foreshortening factor. Explain Isometric, Diametric and Transformations
Trimetric projection using foreshortening factors.
……………………………………………………………………………………
……………………………………………………………………………………
……………………………………………………………………………………
……………………………………………………………………………………
4) Show that for Isometric projection the foreshortening factor along x, y and z-axes
must be 2 / 3 , i.e. fx = fy = fz = 2 / 3
……………………………………………………………………………………
……………………………………………………………………………………
……………………………………………………………………………………
……………………………………………………………………………………
5) Consider a parallel projection with the plane of projection having the normal
(1,0,–1) and passing through the origin O(0,0,0) and having a direction of
projection d = (–1,0,0). Is it orthographic projection? Explain your answer with
reason.
……………………………………………………………………………………
……………………………………………………………………………………
……………………………………………………………………………………
……………………………………………………………………………………
6) Compute the cavalier and cabinet projections with angles of 450 and 300
respectively of a pyramid with a square base of side 4 positioned at the origin in
the xy-plane with a height of 10 along the z-axis.
……………………………………………………………………………………
……………………………………………………………………………………
……………………………………………………………………………………
……………………………………………………………………………………

2.2.2 Perspective Projections

In a perspective projection the center of projection is at finite distance. This projection

is called perspective projection because in this projection faraway objects look small
and nearer objects look bigger. See Figure 15 and 16.

In general, the plane of projection is taken as Z=0 plane.

Properties

1) Faraway objects look smaller.

2) Straight lines are projected to straight lines.
3) Let l1 and l2 be two straight lines parallel to each other. If l1 and l2 are also parallel
to the plane of projection, then the projections of l1 and l2 (call them l’1 and l’2 ),
will also be parallel to each other.
4) If l1 and l2 be two straight lines parallel to each other, but are not parallel to the
plane of projection, then the projections of l1 and l2 (call them l’1 and l’2 ), will
meet in the plane of projection (see Figure 16).
65
 A
Transformations
L1
y B
A’
Object
Center of
Projection B’ P
m L2
Z -Z P’
E Q’ Q
Projectors Center of
-y Projection
Projection Plane
Figure 15 Figure 16

The infinite lines AB and PQ will be projected to the lines A’B’ and P’Q’ respectively
on the plane of projection. That is all points of the line AB is projected to all points of
the line A’B’. Similarly all points of the line PQ is projected to all points of the line
P’Q’. But A’B’ and P’Q’ intersect at M and M is the projection of some point on the
line AB as well as on PQ, but AB║ PQ, which implies that M is the projection of
point at infinity where AB and PQ meet. In this case M is called a Vanishing point.

Principle Vanishing point

Suppose l1 and l2 be two straight lines parallel to each other, which are also parallel to
x-axis. If the projection of l1 and l2 (call them l’1 and l’2 ), appears to meet at a point
(point at infinity), then the point is called a Principle vanishing point w.r.t. the x-axis.
Similarly we have Principle vanishing point w.r.t. the y-axis and z-axis.
Remark
A Perspective projection can have at most 3 Principle Vanishing points and at least
one Principle vanishing point.
To understand the effects of a perspective transformation, consider the Figure 17.
This figure shows the perspective transformation on z=0 plane of a given line AB
which is parallel to the z-axis. The A*B* is the projected line of the given line AB in
the z=0 plane. Let a centre of projection be at (0,0,-d) on the z-axis. The Figure (A)
shows that the line A’B’ intersects the z=0 plane at the same point as the line AB. It
also intersects the z-axis at z=+d. It means the perspective transformation has
transformed the intersection point.
Z
B

Initial line
Vanishing Point
A
B’
d
A’

B*
A*
Projection plane
–d
X

Center of
projection –Z Figure 17

66
 Viewing
Mathematical description of a Perspective Projection Transformations
A perspective transformation is determined by prescribing a C.O.P. and a viewing
plane. Let P(x,y,z) be any object point in 3D and C.O.P. is at E(0,0,-d). The problem
is to determine the image point coordinates P’(x’,y’,z’) on the Z=0 plane (see
Figure 18).
x

P (x, y, z)

.
P’(x’,y’, z’)
z = 0 plane

E (0, 0 – d)
Figure 18
y

The parametric equation of a line EP, starting from E and passing through P is:
E+t(P-E) 0<t<∞
=(0,0,-d)+t[(x,y,z)-(0,0,-d)]
=(0,0,-d)+t(x,y,z+d)
=[t.x, t.y, -d+t.(z+d)]
Point P’ is obtained, when t=t*
That is, P’=(x’,y’,z’) =[t*.x, t*.y, -d+t*.(z+d)]
Since P’ lies on Z=0 plane implies -d+t*.(z+d)=0 must be true, that is t*=d/(z+d) is
true.

Thus x’=t*.x=x.d/(z+d)
y’=t*.y=y.d/(z+d)
z’=-d+t*(z+d)=0

thus P’=( x.d/(z+d), y.d/(z+d), 0)

=(x/((z/d)+1),y/((z/d)+1),0)
in terms of Homogeneous coordinate system P’=(x,y,0,(z/d)+1). ---------(5)

The above equation can be written in matrix form as:

1 0 0 0 
 
0 1 0 0 
P(x’,y’,z’,1)=(x,y,z,1)  = [x,y,0,(z/d)+1] ------------(1)
0 0 0 1/ d 
 
0 0 0 1 

That is, P’h = Ph.Pper,z ----------------- (2)

Where Pper,z in equation (4.6) represents the single point perspective transformation
on z-axis.

The Ordinary coordinates are:

[x’,y’,z’,1]=[x/(r.z+1),y/(r.z+1),0,1] where r=1/d ---------------- (3)
67
Transformations
Vanishing Point

The vanishing point is that point at which parallel lines appear to converge and vanish.
A practical example is a long straight railroad track.

To illustrate this concept, consider the Figure 17 which shows a perspective

transformation onto z=0 plane. The Figure17 shows a Projected line A*B* of given
line AB parallel to the z-axis. The center of projection is at (0,0,-d) and z=0 be the
projection plane.

Consider the perspective transformation of the point at infinity on the +z-axis, i.e.,

1 0 0 0 
 
0 1 0 0 
[0,0,1,0] 0 = (0,0,0,1/d) -------------- (4)
0 0 1/ d 
 
0 0 0 1 


Thus, the ordinary coordinates of a point (x’,y’,z’,1)=(0,0,0,1), corresponding to the

transformed point at infinity on the z-axis, is now a finite point. This means that the
entire semi-infinite positive space(0<=z<=∞) is transformed to the finite positive half
space 0<=z’<=d.

Single point perspective transformation

In order to derive the single point perspective transformations along x and y-axes, we
construct Figures (19) and (20) similar to Figure 18, but with the corresponding
COP’s at E(-d,0,0) and E(0,-d,0) on the negative x and y-axes respectively.

y x

P (x,y,z)
P (x,y,z)

P’(x’,y’,z’) P’(x’,y’,z’)

–x E (–d, 0, 0) E (–d, 0, 0)
y
Figure 19 z Figure 20

The parametric equation of a line EP, starting from E and passing through P is:
E+t(P-E) 0<t<∞
=(-d,0,0)+t[(x,y,z)-(-d,0,0)]
=(-d,0,0)+t[x+d,y,z]
=[-d+t.(x+d), t.y, t.z]
Point P’ is obtained, when t=t*
That is, P’=(x’,y’,z’) =[-d+t*.(x+d), t*.y, t*.z]
Since, P’ lies on X=0 plane implies -d+t*.(x+d)=0 must be true, that is t*=d/(x+d) is
true.

68
 Viewing
Thus, x’=-d+t*(x+d)=0 Transformations
y’=t*.y=y.d/(x+d)
z’=t*.z=z.d/(x+d)
thus P’=( 0, y.d/(z+d), z.d/(x+d))
=(0,y/((z/d)+1), z/((x/d)+1))

in terms of Homogeneous coordinate system P’=(0,y,z,(x/d)+1).

The above equation can be written in matrix form as:

0 0 0 1/ d 
 
0 1 0 0 
P(x’,y’,z’,1)=(x,y,z,1)  =[0,y,z,(x/d)+1]
0 0 1 0 
 
0 0 0 1 

=[0,y/((z/d)+1), z/((x/d)+1),1] -------- (5)

That is, P’h = Ph.Pper,x --------------------------------(6)

Where Pper,z in equation (5) represents the single point perspective transformation
w.r.t. x-axis.

Thus, the ordinary coordinates(projected point P’ of a given point P) of a single point

perspective transformation w.r.t. x-axis is:

(x’,y’,z’,1)= [0,y/((z/d)+1), z/((x/d)+1),1] has a center of projection at [-d,0,0,1] and a

vanishing point located on the x-axis at [0,0,0,1]

Similarly, the single point perspective transformation w.r.t. y-axis is therefore:

1 0 0 0 
 
0 0 0 1/ d 
P(x’,y’,z’,1)=(x,y,z,1)  =[x,0,z,(y/d)+1]
0 0 1 0 
 
0 0 0 1 

=[x/((y/d)+1),0, z/((y/d)+1),1]

That is, P’h = Ph.Pper,y -----------------------------(7)

Where Pper,y in equation (5) represents the single point perspective transformation
w.r.t. y-axis.

Thus, the ordinary coordinates(projected point P’ of a given point P) of a single point

perspective transformation w.r.t. y-axis is:

(x’,y’,z’,1)=[x/((y/d)+1),0, z/((y/d)+1),1] has a center of projection at [0,-d,0,1] and a

vanishing point located on the y-axis at [0,0,0,1].

Example 6: Obtain a transformation matrix for perspective projection for a given

object projected onto x=3 plane as viewed from (5,0,0).

Solution: Plane of projection: x = 3 (given)

Let P (x, y, z) be any point in the space. We know the
Parametric equation of a line AB, starting from A and passing 69
Transformations through B is
y

Projection plane
P (x,y,z)

P’(x’,y’z’)

x
3
E (5, 0, 0)
x = 3 plane

z Figure i

P (t) = A + t. (B – A), o < t < ∞

So that parametric equation of a line starting from E (5,0,0)

and passing through P (x, y, z) is:

E + t ( P – E) , o < t < ∞.
= (5, 0, 0) + t [(x, y, z) – (5, 0, 0)]
= (5, 0, 0) + [t (x – 5), t. y, t. z]
= [t. (x – 5) + 5, t. y, t. z]. Assume

Point P’ is obtained, when t = t*

∴ P’ = (x’, y’, z’) = [t* (x – 5) + 5, t*y, t*. z]

Since, P’ lies on x = 3 plane, so

t* (x – 5) + 5 = 3 must be true;
−2
t* =
x−5
 − 2.y − 2.z 
P’ = (x’, y’, z’) =  3, , 
 x −5 x −5
 3 x − 15 − 2. y − 2.z 
= , , 
 x−5 x −5 x−5

In Homogeneous coordinate system

 3 x − 15 − 2. y 2.z 
P’ = (x’, y’, z’, 1) =  , , ,1
 x−5 x−5 x−5 
= (3x – 15, – 2.y, – 2.z, x – 5) --------------(1)
In Matrix form:
 3 0 0 1
 0 − 2 0 0 
(x’, y’, z’, 1) = (x, y, z, 1)  -------------- (2)
 0 −2 0 0 
 
− 15 0 0 − 5

70
 Viewing
Transformations
Thus, equation (2) is the required transformation matrix for perspective view from
(5, 0, 0).

Example 7: Consider the line segment AB in 3D parallel to the z-axis with end points
A (– 5,4,2) and B (5,-6,18). Perform a perspective projection on the X=0
plane, where the eye is placed at (10,0,10).

Solution: Let P (x, y, z) be any point in the space.

The parametric equation of a line starting from E and passing through P is:

E + t. (P – E), o < t < 1.

= (10,0,10) + t. [(x, y, z) – (10, 0, 10)]
= (10, 0,10) + t [(x – 10)], y (z – 10)]
= (t. (x – 10) + 10, t. y, t (z – 10) + 10)

Assume point P’ can be obtained, when t = t*

∴P’ = (x’, y’, z’) = (t* (x – 10) + 10, t*.y, t*. (z – 10) + 10)

since point P’ lies on x = 0 plane

x=0 y
plane
p (x, y, z)

P’ (x’, y’, z’)

z E (10,0,10)

Figure j

= t* (x – 10) + 10 = 0
− 10
= t* =
x − 10
 − 10.y − 10(z − 10) 
= P’ = (x’, y’, z’) =  0, , + 10 
 x − 10 x − 10 

 − 10.y 10.x − 10.z 

 0, , 
 x − 10 x − 10 

In terms of Homogeneous coordinate system;

 
 − y x − z  x

P’ = (x’, y’, z’, 1) = 0, , ,1 = (0, – y, x – z, − 1)
 x x  10
 −1 −1 
 10 10 

71
Transformations
In Matrix form
0 0 1 1 / 10 
0 −1 0 0 
(x’, y’, z’, 1) = (x, y, z, 1)   ------------------(1)
0 0 −1 −1 
 
0 0 0 −1 

This equation (1) is the required perspective transformation, which gives a coordinates
of a projected point P' (x’, y’, z’) onto the x = 0 plane, when a point p (x, y, z) is
viewed from E (10, 0, 10)

Now, for the given points A (–5, 4, 2) and B (5, –6, 18), A’ and B’ are their projection
on the x = 0 plane.

Then from Equation (1).

0 0 1 1 / 10
0 −1 0 0 
A’ = (x’1’ y’1, z’1, 1) = (– 5, 4, 2, 1). 
0 0 −1 0
 
0 0 0 −1 

−5
= (0, – 4, – 7, – 1)
10
− 15
= (0, – 4, – 7, )
10
= (0, – 40, – 70, – 15)
40 70
= (0, , , 1)
15 15

Hence x1’ = 0 ; y1’ = 2.67 ; z1’ = 4.67

0 0 1 1 / 10
0 −1 0 0 
similarly B’ = (x2’, y2’, z2’, 1) = (5, – 6, 18, 1) . 
0 0 −1 0
 
0 0 0 −1 
= (0, 60, – 130, – 5)

= (0, – 12, 26, 1)

Hence x2’ = 0 ; y2’ = – 12 ; z2’ = 26

Thus the projected points A’ and B’ of a given points A and B are:

A’ = (x1’, y1’z1’) = (0, 2.67, 4.67) and B’ = (x2’, y2’, z2’) = (0, – 12, 26, 1)

Example 8: Consider the line segment AB in Figure k, parallel to the z-axis with end
points A (3, 2, 4) and B (3, 2, 8). Perform a perspective projection onto the z = 0 plane
from the center of projection at E (0, 0, – 2). Also find out the vanishing point.

72
 Viewing
B Transformations

(0,0,2)
Vanishing Point
A y

B’

d=2 A’

B*
A*

d=–2 x

E (0,0, – 2)
Figure k

Solution. We know that (from Equation (1)), the center of single point perspective
transformation: of a point P (x, y, z) onto z = 0 plane, where center of projection is at
(0, 0, – d) is given by:
1 0 0 0
0 1 0 0 
(x’, y’, z’, 1) = (x, y, z, 1) . 
0 0 0 1 / d 
 
0 0 0 1
P’n = Pn. Pper,z ------------------------(I)

Thus the perspective transformation of a given line AB to A* B* with d = 2 is given

by:
V’n = Vn. Pper,z
1 0 0 0
A*  x1* y1* z1* 1 A 3 2 4 1 0 1 0 0 
 * * * =  
B* . 0 0 0.5
 x1 y2 z1 1 B 3 2 8 1 0
 
0 0 0 1
A*  1 0.667 0 1
= 0.6
B*  0.4 0 1

Hence, the projected points of a given line AB is:

A* = (1, 0.667, 0)
B* = (0.6, 0.4, 0)

The vanishing point is (0, 0, 0).

73
Transformations Example 9: Perform a perspective projection onto the z = 0 plane of the unit cube,
shown in Figure (l) from the cop at E (0, 0, 10) on the z-axis.

H
G

D C

E F x
A B

z
Figure (l)
01: Here center of projection
E = (0, 0, –d) = (0, 0, 10).
∴ d = – 10

we know that (from equation – 1), the single point perspective transformation of the
projection with z = 0, plane, where cop is at (0, 0, –d) is given by:

1 0 0 0
0 1 0 0 
(x,’, y,’ z’, 1) = (x, y, z, 1)  -----------(I)
0 0 0 1/ d 
 
0 0 0 1

Pn’ = P. Pper, z --------------- (II)

Thus the perspective transformation of a given cube v = [ABCDEFGH] to V’ =

[A’B’C’D’E’F’G’H’] with d = – 10 is given by:

[V’] = [V] . [Pper, z]

0 0 1 1
1 0 1 1 

1 1 1 1 1 0 0 0
  0
0 1 1 1 1 0 0 
=  . 
0 0 0 1 0 0 0 − 0.1
   
1 0 0 1 0 0 0 1
1 1 0 1
 
0 1 0 1

74
 Viewing
0 0 0 0.9 A'  0 0 0 1 Transformations
1 0 0 0.9 B' 1.11 0 0 1
 

1 1 0 0.9 C' 1.11 1.11 0 1
   
0 1 0 0.9 D'  0 1.11 0 1
V’ =  =
0 0 0 1  E'  0 0 0 1
   
1 0 0 1  F'  1 0 0 1
1 1 0 1  G'  1 1 0 1
   
0 1 0 1  H'  0 1 0 1

Thus the projected points of a given cube V = [ABCDEFGH] are:

A’ =(0, 0, 0), B’ = (1.11, 0, 0), C’ = (1.11, 1.11, 0), D’ = (0, 1.11, 0), E’ = (0, 0, 0)
F’ = (1, 0, 0), G’ = (1, 1, 0) and H’ = (0, 1, 0).

 Check Your Progress 3

1) Obtain the perspective transformation onto z = d plane, where the c. o. p. is at the
origin.
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………

2) Consider a cube given in example – 4, the cube can be centered on the z-axis by
1
translating it – units in the x y directions perform a single point perspective
2
transformation onto the z = 0 plane, with c. o. p. at Zc = 10 on the z-axis.
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………

3) A unit cube is placed at the origin such that its 3-edges are lying along the x, y
and z-axes. The cube is rotated about the y-axis by 30°. Obtain the perspective
projection of the cube viewed from (80, 0, 60) on the z = 0 plane.
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………

Two-Point and Three-Point Perspective transformations

The 2-point perspective projection can be obtained by rotating about one of the
principal axis only and projecting on X=0 (or Y=0 or Z=0) plane. To discuss the
phenomenon practically consider an example for 3-point perspective projection (given
below) some can be done for 2-point aspect.
75
Transformations Example 10: Find the principal vanishing points, when the object is first rotated w.r.t.
the y-axis by – 30° and x-axis by 45°, and projected onto z = 0 plane, with the center
of projection being (0, 0, – 5).

Solution: Rotation about the y-axis with angle of rotation

θ = (– 30°) is
 cos(30°) 0 − sin( −30°)
[Ry] = [Ry]θ = – 30 =  0 1 0 

sin( −30°) 0 cos(−30°) 
 3 2 0 1/ 2 
 
= 0 1 0 
− 1 / 2 0 3 2
 
Similarly Rotation about the x-axis with angle of Rotation φ 45° is:
 
 
1 0 0 
1 1 
[Rx] = [Rx]45° = 0
 2 2
 1 1 
0 − 
 2 2 

 3/2 0 1/ 2  1 0 0 
  0 1 / 2 1 / 2 
∴ [Ry].[Rx] =  0 1 0 .  
 −1 / 2 0 3 / 2 0 − 1 / 2 1 / 2 
 
 3 / 2 −1 / 2 2 1/ 2 2 
 
= 0 1/ 2 1 / 2  -----------------(1)
 −1 / 2 − 3 / 2 2 3 / 2 2
 

Projection: Center of projection is E (0, 0, – 5) and plane of projection is z = 0 plane.

For any point p (x, y, z) from the object, the Equation of the ray starting from E and
passing through the point P is:

E + t (P – E), t > 0
i.e. (0, 0, –5) + t [(x, y, z) – (0, 0, –5)]
= (0, 0, – 5) + t (x, y, z + 5)
= (t.x, t. y, – 5 + t (z + 5)
for this point to be lie on z = 0 plane, we have:
– 5 + t (z + 5) = 0
5
∴t=
z +5
∴ the projection point of p (x, y, z) will be:
 5.x 5. y 
P’ = (x’, y’, z’) =  , ,0 
 z +5 z+5 
In terms of homogeneous coordinates, the projection matrix will become:

5 0 0 0
0 5 0 0
[P] =  -------------(2)
0 0 0 1
 
0 0 0 5

76
 Viewing
 3 / 2 −1/ 2 2 1/ 2 2 0 5 0 0 0 Transformations
  0 5 0 0
0 1/ 2 1/ 2 0
∴[Ry]. [Rx].[P] =  . 
−1 2 − 3 / 2 2 3/2 2 0 0 0 0 1
   
 0 0 0 1 0 0 0 5

5 3 −5 1 
 0 
 2 2 2 2 2
 5 1 
0 0
=  2 2  ----------------(3)
 
 −5 −5 3
0
3 
 2 2 2 2 2
 0 0 0 5 


Let (x, y, z) be projected, under the combined transformation (3) to (x’, y’, z’), then

5 3 −5 1 
 0 
 2 2 2 2 2
 5 1 
0 0
(x’, y’, z’, 1) = (x, y, z, 1)  2 2 
 
 −5 −5 3
0
3 
 2 2 2 2 2
 0 0 0 5 


5 3 5 
 .x − .z 
 2 2 
= x’ = 
 x y 3.z 
 + + + 5
2 2 2 2 2 
 

and
 −5
 5 5 3  
.x + .y − .z 
2 2 2 2 2  
y’ = 
 -----------(4)
 x y 3 
 + + 
.z + 5 
2 2 2 2 2 
 

Case 1: Principal vanishing point w.r.t the x-axis.

By considering first row of the matrix (Equation – (3)), we can claim that the principal
vanishing point (w.r.t) the x-axis) will be:


 5 3 − 5 
 2 , 2 2 ,0 
 1 1 
 
 2 2 2 2 
i.e., (5 6 , – 5, 0) -------------------(I)

In order to varify our claim, consider the line segments AB, CD, which are parallel to
the x-axis, where A = (0, 0, 0), B = (1, 0, 0), C = (1, 1, 0), D = (0, 1, 0)

If A’, B’, C’, D’ are the projections of A, B, C, D, respectively, under the projection
matrix (3), then 77
Transformations
 −5 
 
5 3
A’ = (0, 0, 0), B’ =  , 2 2 ,0 
 1 1 
 +5 +5 
2 2 2 2 
 5 3  5 5  
  − + 
 2  2 2 2 
C’ =  , ,0 
  1 + 1 + 5   1 + 1 + 5  
2 2 2
   
  2 2 2  
 
 
 5/ 2 
D’ =  0, ,0  {Using Equation (4)}
  1 + 5  
  2  
   
 5 6 −5 
 , ,0 ,
A’ = (0,0,0) , B’ =  1 + 10 2 1 + 10 2 


 5 6 5
C’ =  , ,0  and

 3 + 10 2 3 + 10 2 
 5 
D’ =  0, ,0 
 1+ 5 2 

Consider the line equation of A’B’: The parametric Equation is:

A’ + t (B’ – A’)

 5 6 −5 
i.e. (0, 0, 0) + t  , ,0 

 1 + 10 2 1 + 10 2 
 5t 6 − 5.t 
=  , ,0 

 1 + 10 2 1 + 10 2 

we will verify that the vanishing point (I) lies on this line:

 5t 6 − 5.t 
i.e.  , ,0  = (5 6 , – 5,0)
 1 + 10 2 1 + 10 2 
 

5.t. 6
= =5 6
1 + 10 2

−5t
and =–5 -----------------(5)
1 + 10 2

must be true for some ‘t’ value.

t = (1 + 10 2 )

then the equation (5) is true and hence (I) lies on the line A’B’.
Similarly consider the line equation C’D’: The parametric Equation is:

C’ + s (D’ – C’) i.e.

78
 Viewing
 5 6 5   5   5 6 5  Transformations
=  , ,0  + s  0, ,0  −  , 0 
  
 3 + 10 2 3 + 10 2   1 + 5 2   3 + 10 2 3 + 10 2 

 5 6 5   −5 6 5 5 
=  , ,0  + s 

, − ,0  and
 3 + 10 2 1 + 5 2 3 + 10 2 
 3 + 10 2 3 + 10 2   
 5 6 − 5.s. 6 5 5.s.(2 + 5 2) 
= , + ,0  , but
 3 + 10 2 3 + 10 2 3 + 10 2 (1 + 5 2 )(3 + 10 2 ) 
 

we have to verify that the vanishing point (I) lies on C’D’.

i.e. we have to show

 5 6 5  s (2 + 5 2 )  
 (1 − s) 1 + ,0  = (5 6 , – 5, 0)
 3 + 10 2 3 + 10 2  (1 + 5 2 )  
 

for some ‘s’ value This holds true if

5 6
(1 − s ) = 5 6
3 + 10 2
5  s (2 + 5 2 ) 
and 1 +  =–5 ------------(6)
3 + 10 2  (1 + 5 2 ) 

must holds simultaneously for some ‘s’ value.

If we choose s = –2 (1 + 5 2 ), then both the conditions of (6) satisfied

∴(5 6 , –5, 0) lies on C’D’

= (5 6 , – 5, 0) is the point at intersection of A’B’ and C’D’.

(5 6 , – 5, 0) is the principal vanishing point w.r.t. the x-axis.

Case 2: Principal vanishing point w.r.t y-axis:-

From the 2nd Row of the matrix (Equation (3)), the principal vanishing point w.r.t
y-axis will be:

 5 1 
 0, ,0,  in homogeneous system.
 2 2

The vanishing point in Cartesian system is:

 5/ 2 
 0, ,0  = (0, 5, 0) ------------------------(II)
 1/ 2 
 

similar proof can be made to verify our claim:

Case 3: Principal vanishing point w.r.t z-axis:

From the 3rd row of matrix equation (3), we claim that the principal vanishing point
 5 −5 3 3 
w.r.t z-axis will be:  − , ,0, in Homogeneous system.
 2 2 2 2 2 

79
Transformations In Cartesian system, the vanishing point is:
  −5 3  
   
 (−5 / 2)  2 2    −5 2 
  
 , ,0 =  ,−5, 0 ------------------(III)
 3  3    3 
   
 2 2 
 2 2 
   

A similar proof can be made to verify (III)

General Perspective transformation with COP at the origin

Let the given point P(x,y,z) be projected as P’(x’,y’,z’) onto the plane of projection.
The COP is at the origin, denoted by O(0,0,0). Suppose the plane of projection
defined by the normal vector N=n1I+n2J+n3K and passing through the reference point
R0(x0,y0,z0). From Figure 21, the vectors PO and P’O have the same direction. The
vector P’O is a factor of PO. Hence they are related by the equation: P’O = α PO,
comparing components we have x’=α.x y’=α.y z’=α.z we now find the value of α.

R0 (x0, y0, z0)

z
N (n1, n2, n3)

p(x,y,z)

p'(x’,y’,z’)

0 (0, 0, 0) y
x Figure 21

We know that the equation of the projection plane passing through a reference point
R0 and having a normal vector N=n1I+n2J+n3K is given by PR0.N=0, that is

(x-x0,y-y0,z-z0).( n1,n2,n3)=0 i.e. n1.( x-x0)+ n2.( y-y0)+ n3.( z-z0)=0 ---------( )

since P’(x’,y’,z’) lies on this plane, thus we have: n1.( x’-x0)+ n2.( y’-y0)+ n3.( z’-z0)=0
After substituting x’=α.x ; y’=α.y ; z’=α.z, we have :

α =(n1.x0+ n2.y0+ n3.z0)/(n1.x+ n2.y+ n3.z) = d0/(n1.x+ n2.y+ n3.z)

This projection transformation cannot be represented as a 3x3 matrix transformation.

However, by using the homogeneous coordinate representation for 3D, we can write
this projection transformation as:

d0 0 0 n1
0 d0 0 n2
Pper,N,Ro= 0 0 d0 n3
0 0 0 0

Thus, the projected point P’h(x’,y’,z’,1) of given point Ph(x, y, z, 1) can be obtained as

80
 Viewing
d0 0 0 n1  Transformations
 
 0 d0 0 n2 
P’h = Ph. Pper,N, Ro = [x, y, z, 1]  -----------------(16)
0 0 d0 n3 
 
 0 0 0 0 

= [d0.x, d0.y, d0z, (n1.x + n2.y + n3.z)]
Where d0 = n1.x0 + n2.y0 + n3. z0.

General Perspective transformation w.r.t. an arbitrary COP

Let the COP is at C(a,b,c), as shown in Figure 22.

From Figure 7, the vectors CP and CP’ have the same direction. The vector CP’ is a
factor of CP, that is CP’=α.CP

Thus, (x’-a)= α.(x-a) z

(y’-b)= α.(y-b)
(z’-c)= α.(z-c) -------------(17)

N = (n1, n2, n3) z

R0(x0, y0, z0)
N = (n1, n2, n3)

p(x,y,z)

p'(x’,y’,z’)

c (a,b,c)

x Figure 22 y

We know that the projection plane passing through a reference point R0(x0,y0,z0) and
having a normal vector N= n1I+n2J+n3K, satisfies the following equation:

n1.(x-x0)+n2.(y-y0)+n3.(z-z0)=0

Since P’(x’,y’,z’) lies on this plane, we have:

n1.(x’-x0)+n2.(y’-y0)+n3.(z’-z0)=0

Substituting the value of x’, y’ and z’, we have:

α= (n1.(x0-a)+n2.(y0-b)+n3.(z0-c))/( n1.(x-a)+n2.(y-b)+n3.(z-c))
=((n1.x0+n2.y0+n3.z0)-(n1.a+n2.b+n3.c))/(n1.(x-a)+n2.(y-b)+n3.(z-c))
=(d0-d1)/(n1.(x-a)+n2.(y-b)+n3.(z-c))
=d/(n1.(x-a)+n2.(y-b)+n3.(z-c))

Here, d=d0-d1= (n1.x0+n2.y0+n3.z0)- (n1.a+n2.b+n3.c) represents perpendicular

distance from COP, C to the projection plane.

In order to find out the general perspective transformation matrix, we have to proceed
as follows:

Translate COP, C(a,b,c) to the origin. Now, R’0=(x0-a,y0-b,z0-c) becomes the

reference point of the translated plane.(but Normal vector will remain same).

Apply the general perspective transformation Pper,N,R’o

81
Transformations
Translate the origin back to C.

 1 0 0 0 d 0 0 0 1 0 0 0
     
 0 1 0 0 0 d 0 0 0 1 0 0
= 0
 0 0 1 0 0 d 0 0 0 1 0
     
− a − b − c 1  n n2 n3 1  a b c 1 
  1 

 d + n1.a n1.b n1.c n1 

 
 n2 .a d + n2 .b n2 .c n2 
= ------------ (18)
n .a n3 .b d + n3 .c n3 
 3 
 − a.d − b.d 0 − c.d 0 − d1 
 0

Where d = N.CR’ 0 = d0 – d1 = (n1. x0 + n2. Y0 + n3.z0) – (n1.a+n2.b +n3.c)

= n1. (x0 – a) + n2. (y0 – b) + n3. (z0 – c)

And d1 = n1.a + n2.b + n3.c

Example 11: Obtain the perspective transformation onto z = – 2 Plane, where the
center of projection is at (0, 0, 18).

Solution: Here centre of projection, C (a, b, c) = (0, 0, 18)

∴ (n1, n2, n3) = (0, 0, 1)

and Reference point R0 (x0, y0, z0) = (0, 0, – 2)

∴ d0 = (n1x0 + n2.y0 + n3z0) = – 2

d1 = (n1.a + n2.b + n3. c) = 18

we know that the general perspective transformation when cop is not at the origin is
given by:

 d + n1 .a n1 .b n1 .c n1 
 
 n 2 .a d + n 2 .b n 2 .c n2 
 n .a n3 .b d + n3 .c n 3 
 3 
 − a.d − b.d 0 − c.d 0 − d 1 
 0

 − 20 0 0 0   − 20 0 0 0
   
 0 − 20 0 0   0 − 20 0 0
=  0 =
0 −2 1  0 0 −2 1
   
 0 0 
36 − 18   −2 1 
  0 0

Example 12: Find the perspective transformation matrix on to z = 5 plane, when the
c.o.p is at origin.

Solution. Since z = 5 is parallel to z = 0 plane, the normal is the same as the unit
vecter ‘k’.

∴ (n1, n2, n3) = (0, 0, 1)

and the Reference point R0 (x0, y0, z0) = (0, 0, 5)

d0 = n1. x0 + n2. y0 + n3. z0) = 5

82
 Viewing
we know the general perspective transformation, when cop is at origin is given by: Transformations

 d0 0 0 n1   4 0 0 0
   
0 d0 0 n2   0 4 0 0
0 =
0 d0 n3   0 0 4 1
   
0 0 0 0   0 0 0 0 


 Check Your Progress 4

1) Determine the vanishing points for the following perspective transformation
matrix:
8.68 5.6 0 2.8
 0 20.5 0 4.5

 7.0 800 0 2.0
 
 5.3 7.3 0 3.0

………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
2) Find the three-point perspective transformation with vanishing points at Vx = 5, Vy
=5 and Vz = – 5, for a Given eight vertices of a cube A (0, 0, 1), B (1, 0, 1), C (1,
1, 1) D (0, 1, 1), E (0, 0, 0), F (1, 0, 0), G (1, 1, 0), H (0, 1, 0).
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………

2.3 SUMMARY
• Projection is basically a transformation (mapping) of 3D objects on 2D screen.
• Projection is broadly categorised into Parallel and Perspective projections
depending on whether the rays from the object converge at the COP or not.
• If the distance of COP from the projection plane is finite, then we have
Perspective projection. This is called perspective because faraway objects look
smaller and nearer objects look bigger.
• When the distance of COP from the projection plane is infinite, then rays from the
objects become parallel. This type of projection is called parallel projection.
• Parallel projection can be categorised according to the angle that the direction of
projection makes with the projection plane.
• If the direction of projection of rays is perpendicular to the projection plane, we
have an Orthographic projection, otherwise an Oblique projection.
• Orthographic (perpendicular) projection shows only one face of a given object,
i.e., only two dimensions: length and width, whereas Oblique projection shows all
the three dimensions, i.e. length, width and height. Thus, an Oblique projection is
one way to show all three dimensions of an object in a single view.
• In Oblique projection the line perpendicular to the projection plane are
foreshortened (Projected line length is shorter then actual line length) by the
direction of projection of rays. The direction of projection of rays determines the
amount of foreshortening.
• The change in length of the projected line (due to the direction of projection of
rays) is measured in terms of foreshortening factor, f, which is defined as the ratio
of the projected length to its true length.
83
Transformations • In Oblique projection, if foreshortening factor f=1, then we have cavalier
projection and if f=1/2 then cabinet projection.
• The plane of projection may be perpendicular or may not be perpendicular to the
principal axes. If the plane of projection is perpendicular to the principal axes then
we have multiview projection otherwise axonometric projection.
• Depending on the foreshortening factors, we have three different types of
Axonometric projections: Isometric (all foreshortening factors are equal),
Dimetric (any two foreshortening factors equal) and Trimetric (all foreshortening
factors unequal).
• In perspective projection, the parallel lines appear to meet at a point i.e., point at
infinity. This point is called vanishing point. A practical example is a long straight
railroad track, where two parallel railroad tracks appear to meet at infinity.
• A perspective projection can have at most 3 principal vanishing points (points at
infinity w.r.t. x, y, and z-axes, respectively) and at least one principle vanishing
point.
• A single point perspective transformation with the COP along any of the
coordinate axes yields a single vanishing point, where two parallel lines appear to
meet at infinity.
• Two point perspective transformations are obtained by the concatenation of any
two one-point perspective transformations. So we can have 3 two-point
perspective transformations, namely Pper-xy, Pper-yz, Pper-xz .
• Three point perspective transformations can be obtained by the composition of all
the three one-point perspective transformations.

2.4 SOLUTIONS/ANSWERS

Check Your Progress 1

1) Consider a following Figure m, where a given line AB is projected to A’ B’ on a

projection plane.
A

Projector A’

< B
B’
Center of projection

Figure m Plane of projection

a) Center of projection (cop): In case of perspective projection, the rays from an

object converge at the finite point, known as center of projection (cop). In
Figure 1, 0 is the center of projection, where we place our eye to see the projected
image on the view plane.

b) Plane of projection: Projection is basically a mapping of 3D-object on to 2D-

screen. Here 2D-screen, which constitutes the display surface, is known as plane
of projection/view plane. That a plane ( or display surface), where we are
projecting an image of a given 3D-object, is called a plane of projection/view
plane. Figure 1 shows a plane of projection where a given line AB is projected to
A’B’.

84
 Viewing
c) Projector: The mapping of 3D-objects on a view plane are formed by projection Transformations
rays, called the projectors. The intersection of projectors with a view plane form
the projected image of a given 3D-object (see Figure 1).

d) Direction of projection: In case of parallel projection, if the distance of cop from

the projection plane is infinity, then all the rays from the object become parallel
and will have a direction called “direction of projection”. It is denoted by d =
(d1,d2,d3), where d1, d2 and d3 make an angle with positive side of x, y and z axes,
respectively (see Figure n)
y

d = (d1,d2,d3)

z Figure n x
γ
The Categorisation of parallel and perspective projection is based on the fact whether
coming from the object converge at the cop or not. If the rays coming from the object
converges at the centre of projection, then this projection is known as perspective
projection, otherwise parallel projection.

Parallel projection can be categorized into orthographic and Oblique projection.

A parallel projection can be categorized according to the angle that the direction of
projection d makes with the view plane. If d is ⊥ r to the view plane, then this parallel
projection is known as orthographic, otherwise Oblique projection.
Orthographic projection is further subdivided into multiview view plane parallel to the
principal axes)

Axonometric projection (view plane not to the principal axes).

Oblique projection is further subdivided into cavalier

and canbinet and if f = ½ then cabinet projection.
Projection
The Figure O shows the
Taxonomy of projections:

Perspective Parallel

Single-point Two-point Three-point Orthographic Oblique

Multiview Axonometric Cavalier Cabinet

Figure o Isometric Dimetric Trimetric

85
Transformations 3) C

Check Your Progress 2

1) C

2) We know that, the parallel projections can be categorized according to the angle
that the direction of projection d = (d1, d2, d3) makes with the projection plane.
Thus, if direction of projection d is ⊥r to the projection plane then we have
orthographic projection and if the d is not ⊥r to the projection plane then we have
oblique projection.

3) The ratio of projected length of a given line to its true length is called the
foreshortening factor w.r.t. a given direction.
Let AB is any given line segment
r
Also assume AB || a .

Then Under parallel projection, AB is projected to A’B’; The change in the length
of projected line is measured in terms of foreshortening factor. f.
| A' B' |
∴f=
| AB |

Depending on foreshortening factors, we have (3) different types of Axonometric

projections:

• Isometric
• Diametric
• Trimetric

When all foreshortening factors along the x-, y- and z-axes are equal, i.e., fx = fy =
fz, then we have Isometric projection, i.e., the direction of projection makes equal
angle with all the positive sides of x, y, and z-axes, respectively.

Similarly, if any two foreshortening factors are equal, i.e., fx = fy or fy = fz or fx = fz

then, we have Diametric projection. If all the foreshortening factors are unequal d
makes unequal angles with x, y, and z-axes/, then we have Trimetric projection.

4) Refer 2. 3. 1. 2 Isometric projection.

5) For orthographic projection, Normal vector N should be parallel to the direction

of projection vecter, d .
i.e. d = k N where k is a constant.
(-1, 0, 0) = k(1, 0, -1)
This is not possible

Hence, the projection plane is not perpendicular to the direction of projection.

Hence it is not an orthographic projection.

6) The transformation matrix for cavalier and cabinet projections are given by:

 1 0 0 0  1 0 0 0  1 0 0 0
 0 1 0 0  0 1 0 0  0 1 0 0
Pcav =  =  =  -----1)
f.cosθ f.sinθ 0 0   cos 45° sin 45° 0 0  1/ 2 1/ 2 0 0
     
 0 0 0 1  0 0 0 1  0 0 0 1

86
 Viewing
 1 0 0 0  1 0 0 0  1 0 0 0 Transformations
 0    
1 0 0  0 1 0 0 0 1 0 0
&Pcab=  = 1 1 = ---(2)
f.cosθ f.sinθ 0 0  . sin 30° . cos 30° 0 0 0.43 0.25 0 0
  2 2   
 0 0 0 1  0 0 0 1  0 0 0 1

The given pyromid can be shown by the following Figure p.

y
E
–z

A D

o –x
x
B C
Figure p
z
The vertices of the pyramid are:

A (2, 0, – 2), B (2, 0, 2), C (– 2, 0, 2)

D (– 2, 0, – 2), E (0, 10, 0)

Using the projection matrices from (1) and (2), we can easily compute the new
vertices of the pyramid for cavalier and cabinet projections. (refer Example 4).

Check Your Progress 3

1) Let p (x, y, z) be any point in 3D and the cop is E (0, 0, 0).

The parametric equation of the ray, starting from E and passing through p is:

p (x, y, z)

p'(x’,y’,z’)

COP

E (0, 0, 0)
z = d plane –z

y
Figure q

E + t (P – E), t > 0
= (0, 0, 0) + t [(x, y, z) – (0, 0, 0)]
= (t.x, t. y, t. z)

87
Transformations For this projected point of p (x, y, z) will be:
t. z = d
d
= t= must be true.
z

Hence the projected point of p (x, y, z) will be:

 d.x d.y   dx dy 
P’ = (x’, y’, z’) =  , , d  ⇒ in homogenous Coordinates  , , d,1
 z z   z z 
= (dx, dy, dz, z)

In matrix form:
d 0 0 0
0 d 0 0
(x’, y’, z’, 1) = (x, y, z, 1) 
0 0 d 1
 
0 0 0 0

2) Since the cube is first translated by –0.5 units in the x and y-directions, to get the
centred cube on the z-axis.

The transformation matrix for translation is:

 1 0 0 0
 0 1 0 0
[Tx, y] =  ----------(1)
 0 0 1 0
 
− 0.5 − 0.5 0 1

A single-point perspective transformation onto the z = 0 plane is given by:

1 0 0 0 
0 1 0 0 
Pper,z =  ----------(2)
0 0 0 1 / d
 
0 0 0 1 

1
It has a center of projection on the z-axis: at d = – 10 ⇒ = – 0.1
d

From equation (2)

1 0 0 0 
0 1 0 0 
Pper,z = 
0 0 0 − 0.1
 
0 0 0 1 

The resulting transformation can be obtained as:

 1 0 0 0 
 0 1 0 0 
[T] = [Tx,y]. [Pper,z] = 
 0 0 0 − 0.1
 
− 0.5 − 0.5 0 1 

88
 Viewing
Thus, the projected points of the centred cube V = [ABCDEFGH] will be: Transformations

0 0 1 1
1 0 1 1 
 .
1 1 1 1  1 0 0 0 
  
0 1 1 1 0 1 0 0 
[V’] = [V]. [T] =  
. 
0 0 0 1  0 0 0 − 0.1
   
1 0 0 1 − 0.5 − 0.5 0 1 
1 1 0 1
 
0 1 0 1 

− 0.5 − 0.5 0 0.9 A'  − 0.56 − 0.56 0 1

 0.5 
− 0.5 0 0.9 
B'  0.56 − 0.56 0 1


 0.5 0.5 0 0.9 C'  0.56 0.56 0 1
   
− 0.5 0.5 0 0.9 D'  − 0.56 − 0.56 0 1
=  =
− 0.5 − 0.5 0 1  E'  − 0.5 − 0.5 0 1
   
 0.5 − 0.5 0 1  F'  0.5 − 0.5 0 1
 0.5 0.5 0 1  G'  0.5 0.5 0 1
   
− 0.5 0.5 0 1  H'  − 0.5 0.5 0 1

3) A unit cube is placed at the origin such that its 3-edges are lying along the x,y,
and z-axes. The cube is rotated about the y-axis by 30°. Obtain the perspective
projection of the cube viewed from (80, 0, 60) on the z= 0 plane.

3) Rotation of a cube by 30° along y-axis,

y
G

H
D C

F
E
(0,0,0)

A B

Figure r
z

cos 30° 0 − sin 30° 0

 0 1 0 0
[Ry]30° = 
 sin 30° 0 cos 30° 0
 
 0 0 0 1

89
Transformations  3/2 0 − 1 / 2 0 0.86 0 − 0.5 0
   0
0 1 0 0 1 0 0
=  = 
1/ 2 0 3 / 2 0  0.5 0 8.6 0
   
 0 0 0 1  0 0 0 1

Let p (x, y, z) be any point of a cube in a space and p’ (x’, y’, z’) is its projected point
onto z = 0 plane.

The pametric equation of a line, starting from E (80, 0, 60) and passing through P (x,
y, z) is:
E + t ( P – E), 0 < t ∞.
= (80, 0, 60) + t [(x, y, z) – (80, 0, 60)]
= (80, 0, 60) + t[ (x – 80), y, (z – 60)]
= [t (x – 80) + 80, t.y, t. (z – 60) + 60]

Assume point P’ can be obtained, when t = t*

⇒ P’ = (x’, y’, z’) = [t* (x – 80) + 80, t*.y, t* (z – 60) + 60]
Since point p’ lies on z = 0 plane, so
−60
t* (z – 60) + 60 = 0 ⇒ t* =
z − 60
 − 60.x + 80.z − 60. y 
⇒ p’ = (x’, y’, z’) =  , ,0 
 z − 60 z − 60 
In Homogeneous coordinates system:

 − 60.x + 80.z − 60.y 

P’ (x’, y’, z’, 1) =  , ,0,1
 z − 60 z − 60 
= ( – 60. x + 80. z, – 60. Y, 0, z – 60)
In Matrix form:
− 60 0 0 0 
 0 − 60 0 0 
(x’, y’, z’, 1) = (x, y, z, 1).  ------------(1)
 80 0 0 1 
 
 0 0 0 − 60
P’n = Pn. Ppar, z -------------(2)

Since a given cube is rotated about y-axis by 30°, so the final projected point p’ (of a
cube on z = 0 plane) can be obtained as follows:

P’n = Pn. [Ry]30°. Ppar, z

0.86 0 − 0.5 0 − 60 0 0 0 
 0 1 0  
0  0 − 60 0 0 
(x’, y’, z’, 1) = (x, y, z, 1) .  .
 0.5 0 0.86 0   80 0 0 1 
   
 0 0 0 1  0 0 0 − 60

91.9 0 0 − 0.5
 0 − 60 0 0 
(x’, y’, z’, 1) = (x, y, z, 1)  -------------(3)
38.8 0 0 0.86 
 
 0 0 0 − 60 

P”n = P. Ppar, z, 30°

This equation (3) is the required perspective transformation. Which gives a

coordinates of a projected point P’ (x’, y’, z’) onto the z = 0 plane, when a point P (x,
y, z) is viewed from E (80, 0, 60).
90
 Viewing
Thus, all the projected points of a given cube can be obtained as follows: Transformations

A 0 0 1 1
B 1 0 1 1 
C 1 1 1 1 90.9 0 0 − 0.5
   0
D 1 1 1 1  − 60 0 0 
P’ = V. Ppar, z, 30° =
E 0 1 1 1 38.8 0 0 0.86 
   
F 0 0 0 1  0 0 0 − 60 
G 1 0 0 1
 
H 0 1 0 1

A'  38.8 0 0 − 59.14 A'  − 0.72 0 0 1

B' 129.7 0 0 − 59.64 
B' − 2.17 0 0 1

C' 129.7 − 60 0 − 59.64 C' − 2.17 1.01 0 1
   
D'  38.8 − 60 0 − 60.86 D'  − 0.64 0.99 0 1
=  0
=
E' 0 0 − 60.0  E'  0 0 0 0
   
F' 90.9 0 0 − 60.5  F' − 1.50 0 0 1
G' 90.9 − 60 0 − 60.5  
G' − 1.50 0.99 0 1
   
H'  0 − 60 0 − 60.0  H'  0 1 0 1

Hence, A’ = (– 0.72, 0, 0), B’ = (– 2.17, 0, 0), C’ = (– 2.17, 1.01, 0)

D’ = (– 0.64, 0.99, 0), E’ = (0, 0, 0), F’ = (– 1.5, 0, 0)
G’ = (– 1.50, 0.99, 0) and H’ = (0, 1, 0).

Check Your Progress 4

1) The given perspective transformation matrix can be written as:

From Rows one, two and three from equation matrix (I), the vanishing point w.r.t.
x, y and z axis, will be:

Cx = (3.1, 2.0, 0)
Cy = (0, 4.56, 0)
Cz = (3.5, 4.0, 0)

2) From the given V.P., we can obtain the corresponding center of projections. Since
vanishing points: Vx = 5, Vy= 5 and Vz = – 5, hence center of projections is at:

Cx = – 5, Cy = – 5 and Cz = 5

1 1 1 1 −1
∴1/d1 = = 0.2, = = 0.2 and = = – 0.2
5 d2 5 d3 5
Hence, the 3 – point perspective transformation is:

1 0 0 0.2 
0 1 0 0.2
Pper–xyz =  ------------(I)
0 0 1 − 0.2 
 
0 0 0 1 

Thus by multiplying v = [ABCDEFGH] with projection matrix (I), we can obtain the
transformed vertices of a given cube.

91