TWO-DIMENSIONAL | TRANSFORMATIONS hea 1, TRANSFORMATIONS AND. MATRICES __ = Something else by applying | Transformation means changing some graphics into ae ar belle oor rules, We can have various types of transformations such asf trans ee | | 2D it is called” dwn, rotation, shearing, jetc. When a transformation takes place on a 2D plane, = 2D transformation. a Transformations play an important role in computer graphics to reposition the graphics on the screen and change their size or orientation. : One of the most common and important tasks in computer graphics is to transform the coordinates (position, orientation, and size) of either objects within the graphical scene~ or.the camera that is viewing the scene. It is also frequently necessary to transform coordinates from one coordinate system to another, (e.g. world coordinates to viewpoint coordinates to screen coordinates.) All of these transformations can be efficiently and sufficiently handled using some simple matrix representations, which we will see can be Particularly useful for combining multiple transformations into a single composite transform matrix. We will look first at simple translation, scaling, ults to 3D, and finally see how multiple transform, mposite transform. 2._ TRANSLATION IN 2D (Reposition ] Point (X, Y) is to be translated apn Dx and Dy to anew location (x', se X' = Dx +X as ' Y =Dy+Y eos geen 4 = distane orP=T+Pwhere _ and rotation in 2D, then extend our ‘ations can be easily combined into a P = [X > UL Y Dx A Hin, ‘ope , Dy detinati- || Ae AT 524g rat Tx : ' WY) hy 5 y= ¥4 Ty ’ fay 3. SCALING IN2D 7) re r : eam (ry. fy mi? wr Point (X,Y) is to be scaled by n : rn: ¥ amount Sx and Sy to location OY) Y' =Sy*yor P= $"P where Qe P = |x’ Iv |Sx 0| P = [X| A B [0 sy| ‘ C [ l \¥| odsy | | scaling is performed about the pe hs 0,0) not about ¢ jine/polygon/ whatever -SY+0 OFY-SY (2" y] Scale > 1 enlarge the object and ntove it away from thé origin!" Scale = 1 leave the object alone Scale < 1 shrink the object and move it towards the origin. he center of the uniform scaling : Sx = Sy differential scaling Sx != Sy -> alters proportions ROTATION IN 2D J Point (X, Y) is to be rotated about the origin by angle theta to location (X', Y). * cos(theta) - ¥ * sin(thet in(theta) + Y *cos(theta) multiplication * orP'=R*P where’ = | Cos (theta) - Sin (theta) | = | Sin (theta) ~ Cos (theta)| = |x| | = ela rotation is performed about the origin (0,0) not about the center of the line/polygon/ whatever. | bint re 5. HOMOGENEOUS COORDINATES AND MATRIX REPRESENTATION OF 2D) ‘ANSFORMATIONS i /e have seen that basic transformations can be expressed in matrix form. But many 8raphic application involve sequences of geometric transformations. Hence we need a Beneral form of matrix to represent such transformations. This can be expressed as : Pe Pet, +T, Where and P’ — represent the row vectors. T, ~is a2 by 2 array containing multiplicative factors. T, ~is a2 element row matrix containing translation terms. Ze can combine multiplicative and translational terms for 2D geometric formations into a single matrix representation by expanding the 2 by 2 matrixer on (5, ¥ BP vrw Computer Graphics & Animation 5 like Matric wie’ Hue re dOF ind ye done he wwe OF matrioes in computer graphics is widespread hone 1100 a ales erebiteetar, cartoon, automotive that were formerly done by Nand 6 ' pa oon . noatinely with the aid of computer graphies, Video gaming indus ry, ma lndtustey to rely heavily on computer graphics, iy now representing Somensions, eonatmuct and MW vicleo gaming industry, matrices are major math ‘es of matt ee Geld Manipulate a ralistic animation of a polygonal figure, Example ion concepts ike fi Wranslations rotations, and seating, Other matrix tsénaform ations cof Mate’ 18 6 Of view, rendering, color transformation and projection, Understa aslo newessity to program 3D video games * NOMOGENFOUS COORDINATES finates is to capture the One of the many purposes of using homogencous coord Tamathing that does nol SQeceptof infinity, Th the Euclidean éSardinate system, in nt MEOH Bae ‘computations Sxist, Mathematicians have discovered that many geoiielrié "nig Win bacome very clear San be groally SiNpIME inthe concept of infinity is used, T ia act ange eteocs when We HOW"W CUE and surfaces design, Without ne engi aii el SSoninates system, it would be difficult to design certain classes of very and surtaces in computer graphics and computer-aided design. 9 i, Let us consider two real numbers, a and w, and compute the value of a/w. Let us ho} the Value of @ Fiat ane-vary-the Value ow. As w getting smaller, the value of a/w i va Setting larger. If w approaches zero, alw approaches to infinity! to capture tl Soncept of infinity, we use two numbers a and w to represent a value v, v=a/w. if w is not zero, the value is exactly a/w. Otherwise, we identify the infinite value with (4,0) Therefore, the concept of infinity can be represented with a number pair like (a, w) or as a quotient af Let us apply this to the xy-coordinate plane. If we replace x and ¥ with x/w and y/w, a function Axy) = 0 becomes flx/w,y/w) = 0. If function fix,y) = 0is a polynomial, multiplying ‘with w" will clear all denominators, where 1 is the degree of the polynomial. For example, suppose we have a line Ax + By+C=0. Replacing x and Y with x/w and wie yields A(x/w) + Biyw) + C = 0, Multiplying by w changes it to : Ar + By + Cw =0. 4 second degree polynomial Ax2 + 2Bry + Cy2 + 2px - ing x and y with x/w and y/w and multiplying the Bi with w*, 2 + 2Bry + Cy? + 2Dxw Hf you look at these tw. terms are equal. In the ca second degree polynomial, + 2Eyw + Fw? =0 ‘© polynomials carefully, se of a line, terms x, y an | all terms (ie, x2, you will see th. id w are of deg *v, ¥?, xw, yw and w2) a at the degrees of all ree one, while in the re of degree two. Given @ polynomial of degree n, after introducing w, all terms , 2 are of degree © Consequently, these polynomials are called homogeneous i coordi +920) the homogeneous coordinates "zeneous Polynomials and the ae degree » polynomial in a homoge i “ weeneous coordinate system, dividing the polves mia wh ie" and replacing x/w, yw with + and y, respectively, will ae the ymial back to a conventional one e ., . ; rial isthe following, "4! OM: For example, ifthe piner degree 3 homogeneous x + Say? ~ Sy2w + 109 the result is 0 P + Sry -5y2 + 10<0 This works for three-din mn as w) and multiply the result by w M. One can sed to certair replace a point (x, v. 2) a power. The resulting pe2 be exp6 i. wow WN OF THE UNIT SQUARE \ysibta car newer the types nn coommppunet $F2QPICS. Vig 2 Hist SoneOre Tih be used { TRANSHORMATIC ill see how Jig “oh qseatiOn In thin projeet, won't do any pron was programming compys rt in two dimensions, Ir but you could u» ing, a really 3 quare, wetiave we corners in counite 1 the corners of the square are 1S enix soulates everything we need to kn ar transformation ti he vertices, This torr ing, but we will ter graph would need to be able to an are animating, “4 es to arammau Let's ota pretty boring example, We'll Tearn the Hniqui matrix A represe nly yrner of th p mF In typical mathematical fashion, ve put " but you could choose any ones Koowing. wh ze fo reconstruct the qt 0 this A sulficie paula 3 Mbout the unit square. More importantly, when we apply Jine ona ; he transformation to each of t Unit square, its enough to apply the transfor x from the fact that linear transformations take lines to Tir 1. Now let's say in our animation, we want to enlarge t length _ : 2. Find a matrix T 90 that TA isa 2% 4 matrix representing, the enlarges — Fiven though we're doing matrix multiplication, what we are ay oe = computing T x for four different vectors , the four columns of A. Since each colum nts a-corner of the square, we are finding, the image of each of the corners of nder the linear transformation that doubles lengths. Since linear transformation ame techniqu the unit qu fo1100011) rclockwise orientation he square 90 that it has side aqui take Tinea to Tines, we know that By finding the image of each of the four comers, w simply connect them and get the image of the whole square under the linear (ransformation, TU would be unteanonable, not_to mention impossible, if we had « rm cach of the points along, the sides of the square. Transforming the corner ehough. Since our goal is lo compute T * for four different x vectors, we can put the four veetors ina matrix A and compute T A, which is what is typically done. move the square three-units to the right. Try for a little while to find 2 that TA is the 2» 4 matrix representing the shifted square. If you've failed 2 4k, you've succeeded! Show that the transformation T( [xy] )= [x +3 y +3] isnot ar transformation, Since linear (ransformations are exactly matrix transformations ~ we could not have succeeded in finding a matrix to represent this transformation. The fac~ that translation is not a linear transformation seems like ar Computer graphics, How How fun is it to transl should be a major handicap fo" nimated movie be if the characters cannot move? hoot down a bunch of stationary targets in a video game? Since absolutely essential to applications of computer grapics, we need a wa! wind. this problem. The answer lies in homogencous coordinates Inetesd of ing. point on the xy-plane as a vector in R’2, represent it as a vector R 4 where the third coordinate is one, So from now on we will use A= 011000111111. This nt actually allows us to represent translation by -w-mvatrix, this ct - can an jon y trivial adjust 9, “SOLID BODY TRANSFORMATIONS a In_computer graphics there are i ds of ‘itso price there are many different kinds of transformation which V fotation, seale, similarity (refle ation | Rigid traiteforr tr ion) affine, homograph: ri combination of rota Brapiy and rm Rigid body transformations are the object Le. magnitude and the transformation. Unit formation y mal ones which preserve th also. Py we shape and size of th nile also, Pure rolatlons and pure tllectionc ney ac the ange 4 reflections are rigid bod’ ; "6's not a rigid body transformation as Tt eaves themations or A rigid transformation is formally defined as a transformation that, when tor v, produces a transformed vector T(v) of the form acting, or any vec Tw)=Rvtt , vg, the where R' = RB (ie, R is an orthogonal transformation), and t is a vector i translation of the origin A proper rigid transformation has, in additio det(R) = 1 ents a rotation which means that R does not produce a reflection, and hence it represents a rot (an orientation-preserving orthogonal transformation). Indeed, when an orthogonal transformation matrix produces a reflection, its determinant is —1 fi, WINDOW-TO-VIEWPORT MAPPING Definitions @ World Coordinate System : This is object space or the space in which the \ application model is defined. — @ Screen Coordinate System : The space in which the image is displayed. @ World Window (or clipping) : This is the rectangle in the world defining the \ region that is to be displayed @ Interface Window : The window opened on the raster graphics screen in which the image will be displayed. ( (xel\ @ Viewport : The rectangular portion of the interface window that defines where ~ the image will actually appear (usually the entire interface window but in some cases modified to be a portion of the interface window). @ Viewing Transformation : The process of mapping a world window in World Coordinates to the Viewport er a 11. DESCRIPTION Window-to-Viewport mapping is the process of mapping or transforming a two dimensional, world-Coordinate scene f0 device coordinates. In particular, objects inside the World or clipping window are mapped to the viewport. The viewport is disptayed in the interface window on the screen. In other words, the clipping window is used to select | the part of the scene that is to be displayed. The viewport then positions the scene on the Sutput device. ample Fig. 3 Viewport Coordinates YW.mas, nave : \ Competes € " . a « owe Mappin maton To Viewport MARE formation: vat start yells pO, “ granatormation INVOLVES LenS prectice eorronponel ing pole This maps ES wh The formula ts use he proportional’ x spiny KY BO “PROpOrtional ay he wot wise, SEIN lt ihe fo TR TET RC BTS AE of the way (oom th hedge of Ar yw le 90% of the Way> sense that it xw IS vet, Si A) T he way fron the Tet exlge of the viewport AONE GE tHe way Front the Lotte » ™~ ” s . earreterny, GT VEU Fram the tovtom axige of the worl er inavre ie prapartionallly atthe WH te the viewport, The picture below shows 8 ferornernaia i worm 30m cn liesiciaiione —— xvmax aw . Fer proportionality inv: Yvmax ywmex w > — e Ywmin Y¥min. Fig. 5 Using this proportionality, the following ratios must be equal. XV-XV min XW- XW min Vmax XV min ~ XWmax-XWmin .o YV-YV min yW-YWimin — WWmax-YV min ~ max~YWmin By solving these tions for INKNO} oo -_ the unknown viewport position (xv, yw), the follov® XV = SxXW + tr YV = SyYWety The scale factors (Sx, Sy) would be Sx XVmin in = Pyme- YVmin 9 > YWinin And the translation factors (1, T po taedl Y) would be pai XWinin - XVinayvow “1 The_position of the viewport can be changed allowing objects to be viewed at ions onthe Interface Window Multiple viewports can also be used to different sections of a Scene at different screen positions, Also, by changing the ite TRuS: Foote Size and proportions of the objects being displayed ct_can be achieved by succes: sively mapping different ona fixed sized viewport If the aspect ratio of the world window may look distorted. 1 be manipulated. Thus, a zooming affe simensioned clipping windows and the viewport are different, then the image OpenGL Window-To-Viewport Mapping In OpenGL. the function void gluOrtho2D(left, right, bottom, top); is used to set up the world window, For example, if we wanted a wasld window with © vas ying from - 1.0 0 and y varying from 3.0 to 5. 0, we would use the follow ing code to accomplish this glMatrixMode(GL_PROJECTION); glLoadidentity(; gluOrtho2D(-1.0, 1.0, 3.0, 5.0); To set up the viewport, we would use glViewport(left, bottom, width, height); For example, if we wished a viewport to start at the lower left corner of the interface window, have a width of 200 and a height of 300, we would use the following OpenGL statement to accomplish this: glViewport(0, 0, 200, 300); Tiling using the Window-To-Viewport transformation If we draw a number of copies of the same image in rows and columns across the interface window so that they cover the entire window, it is called tiling. The picture that s drawn many times is called a motif. To achieve tiling in computer graphics, the window remains static and the viewport is changed many times and the picture is redrawn each time the viewport is changed. The following picture shows the same image from the world drawn 4 times in 4 different viewports. Fig. 6 This can be achieved using, the following code for (int j 0; 1 <2; i++) for (int j = 0; < 2; j++) | ; | "Wiewport (i * screenWidth/2.0, j* screenHeight/2.0, screenWidth/2, ' *steenHeight/2), | drawSaturn(); Y. ®Sc-.T-Computer Graphics & Animation (Sem.-1V) 4s,2 wee sformation cing Window-to-Viewport tran Ne ie of the area to be viewed iy Zooming or Pan . FT Zooming is a tee ) use detail or le different parts graphics by aken from wikipedia.org) nique in whict can change the detail, Panning or roaming mean cof it at different time leaving the vie liding the camer order to see mor Both zooming, ang port static ar hows these across the scene seaming can be achieved in_ computer nena + window) The picture below (t changing the concepts Fane Crackers 4 frend = | Milk Potatoes I Panning across objects In Zooming out for an {wo directions overview Zooming in for greater ‘detail Fig Reflection through an Arbitrary Line formations of objects. Distortion that can be achieved b P rol points, distortion transformations will change the Poets. The other is affine transformation which preserves the hape of objects. The most important of affine transformations are about a point, reflection about a line and shearing (distortion two types of tra yints and cont ‘ersal tools for computer graphics programs are the matrices that manipulate ectors. Each point or vector on a plane is represented by a single column this topic we will present a direct method how to apply affine reflection through an arbitrary line (y = ax + b) in a mathematica sordinates using analytic geometry REFLECTION THROUGH an ARBITRARY LINE : y = ax +b Geflectons are the simplest operation of transformations. Simple reflections w the object through the x-axis, y-axis, and the origin. Reflection through an arbitrary , « b) is much more complicated and requires some mathematical calculations. a Q yeaxth / C Vpnal Transformations i m Reflection about an Arbitrary Line Using Matrices This can be done using @ combination of other transformation ‘operations: translate, rotate and reflect 1. Translate the Cartesian coordinates so th. at the arbitrary line y = ax + b will pass through the origin O. Rotate the Cartesian coordinates so that one of the coordinate (x-coordinate is preferable) lies onto the arbitrary line y =ax +b. 3. Perform the reflection about the aligned axis (also y = ax + b), 4. Restore the original Carte translation transformations. [2] N coordinate by applying inverse rotation, and This algorithm is easy to understand, but implementation requires several matrices operations. This process is complicated and time consuming in term of executing efficiency. We also need to find the angle of rotation so the x-axis (or y-axis) to lie on the line y = ax + b (step 2). And then after Performs the reflection, we will have to find the inverse matrices to move the coordinates back to its original position (step 4). Algorithm to Find Point Q Reflection of Point P through Line y=ax+b This is a direct approach, using analytic geometry te chnique (Fig. 8). 1. Find equation of line (PQ) that passes through point P and perpendicular to the line y = ax +b. 2. Find the distance (d) from point P to the line y = ax +b. 3. Coordinates of point Q(xq, yq) should satisfied equations of line (PQ) (from #1) and the distance d(q) is also equal the distance from d(p) ot : d(q) = d(p) = d (from #2) Rotation About an Arbitrary Point { Transformation means change in image. We can modify the image by performing some basic transformation such as : © Scaling ao. @ Rotation @ Translation Rotation i ~~ For rotation we need trigonometry logic. Suppose we have point P1 = (xl, y1) and we Totate it about the original by an angle 0 to get a new position P2 = (x2, y2) as shown in figure.rer 4 tit, Translation : Translate back to original po Ss a ix T2 will [ meee a) at ' ” Xm Yu Th : 2+ x ~ & —j—__}-_}_}-__} + >= 2 eS. 8 3 Sew . Fig. 13 combined matrix * Rotation * Translation fr 1 © 0) fose@ smO807 7,1 0 0} =| @ 238) ° sa eoee] 0 at -Xm -Yr 1 ® o 1) (meyen pie ey cos@ sine 0 = 0 10 “|e oes] | -Xm -Ym 1] Xm Ym 1 cos 8 sin 0 = sin 8 cos 8 0 -Xm*cos6+Ym‘sin@+Xm -Xm*sin@-Ym*cos@+¥m 1 This transformation matrix is the overall transformation matrix for rotation about arbitrary point (Xm, Ym) by an angle 6 in anticlockwise direction. THREE-DIMENSIONAL TRANSFORMATIONS ten 2c oa gee We now tum to transformations in three dimensions. In most cases, the mai Se eee ey “in two dimensions, there is only one Cartesian coordinate system, with two (eg aeerrenbtiaais Ghar ant y cael , the axes don’t have to be perpendicular, but this i inelevant for Our discussion ela oe aT a nad bat he ne ‘system is. positive x axis and your index finger with the positive ¥ srl hen pola the direction Of postive =. The le fora Te-handed uses the left hand in a similar manner. It is also possible to define a left-handed Soordinate system as the mirror image (reflection) of a right-handed one. Notice that one | ¥ L ie Projection Plane Fig. 20 Vanishing Points ‘A vanishing point, or point of convergence, is a key element in many works of art. In a linear perspective drawing, the vanishing point is the spot on the horizon line to which the receding, parallel lines diminish. It is what allows us to create drawings, paintings, and photographs that have a three-dimensional look.) The easiest way to illustrate this in real life is to stand in the NN middle of a straight road. When you do this, you'll notice how the sides of the road and the lines painted on it meet in one spot on the horizon. The > center line will go straight for it and the lines on the side will \—~ ——— angle in until all of them intersect. That point of intersection is the vanishing point Every object extends outward in various directions. The fact is, a vanishing point only represents a direction an object 3 points away from cS) Nae: 21 For that reason,“it should just be called a direction point instead. ) (One might even think of the direction point as the direction that something poin ts) As such,(everything going in the same direction will converge to the same direction point. Note how each side of this square goes off toward one of two different directions, each with its own direction point. (left) (Despite all of the complex definitions in Fig. 22 every how to draw tutorial I have seen about the vanishing point, Ihave never come across a single one which stated that simple fact) After its cisco sy Ace artists only understood how to use one vanishing point near the center of their picture which then came to represent where everything is moving away from us. (right) Even when more vanishing points came into use later, this idea of it having a depth and distance-related function has persisted. At best, any other how to draw tutorial will define the vanishing point as being : "That point at which parallel lines appear to converge into the distance.” (top pic) But the only relevance of those parallel lines, which are mentioned in every single Vanishing point definition, is that parallel lines point in the same direction, After all, that's what parallel means. This was the only hint that direction was a part of the process, with any emphasis on it having been missed and placed on distance instead.= evr Onmnnte \ ) But really, (every lativicioal dirwetion om an object has a dirwetion point somewhery and SO we do HOt Het wo oF More parallel lines for these points to exist, (below) Stil, Knowing this is rarely useful, giving parallel lies nvore emphasis, ) whee W) — Converging liv ave the separate funetion of outlining forms, yet are always part of the ranting poet's converging lines a our PTS The subject of *) peotive” is based on using converging lines to sug (top pic) As it tums out, the conventional definition of V really much more than a simple perspective! mee these are two separate things: being assigned to one dot, the vanishing point should just be called a “direction Point! with those converging parallel lines given their own name. Guide lines is he simplest and most descriptive term used, being that they guide a form's directions and shape. definition in every how to draw tutorial, But without those hing point has nothing vanishing” there to justity its name, yest depth, anishing point stated earlier is dot, It is, in tact, describing that overall system of linear Next, observe that a direction point functior equally well with objects that are all lined up in one. direction. (left) This direction point takes into account, not the directions of each object, but all of those objects as a whole. Coordinate Transformations Fig. 24 We can also achieve the effects of translation, scaling , and rotation by moving the observer who views the object and by keeping the object Stationary. This type of ‘ransformation is called a coordinate transformation. We first attach a coordinate system to the observer and then move the observer and the attached coordinate system. Next, we recalculate the coordinates of the observed object with Tespect to this new observer coordinate system. The new coordinate values will be exactly the same as if the observer had remained stationary and the object had moved, corresponding to a geometric transformation (see Fig. 25). PG, e o Gey Pig. 25vs mputer Graphics & Animation Wv > ero ts P (x y, 2) having the homogeneoy ) points ons are then applied to F These transformations ar form x y z 1 Projections » are converted to viewing Once world-coordinate descriptions of the objects in a Rann a a Apaphrt coordinates, we can project the three-dimensional objects ety PropecHeAPeRCRInar lane. There are two basic projection methods. In a pa : RHC Van AAAS le eaten are transformed to the view plane along F rele) i ae praia eel te of Fig. 26. For a perspective projection (Fig. 27), object ie Geek nero et view plane along lines that converge to a point called the Pp oe ied by calculating the center of projection). The projected view of an object is dete intersection of the projection lines with the viewplane. Projectors Centre of Projection Fig, 27 : Perspective projection of an object to the view plane 4 parallel projection preserves relative proportions of objects, and this is the method used in drafting to produce scale drawings of three-dimensional objects. Accurate views of the various sides of an object are obtained with a Parallel projection, but this does not Bive us a realistic representation of the appearance of a three-dimensional object. A eoata ve Projection, on the other hand, produces realistic views but does not preserve relative proportions. Projections of distant objects are smaller than the projections of objects of the same size that are closer to the projection plane Parallel Projections Orthographic. The projection of axis onto a viewplane cont co-ordinate to zerc 4 point by aining, the othe © projector parallel to the ith co-ordinate T two axes is obtained by setting the Itha Hence to project in the direction of the z-axis onto the xy plane can be carried out by Scale (1.0.1.0,0.0) 2. Axonometric, The required transformation is produced in two stages, firstly the direction of projection is rotated until it aligns with one of the co-ordinate axes and then an orthographic projection along that axis is carried out. For example in the isometric case the direction of projection must be symmetric with respect to the three co-ordinate directions to allow equal foreshortening on each axis. This * is obviously achieved by using a direction of projection (1,1,1). 3. Oblique. It is usual to take the viewplane parallel to a face of the object. This face (and these parallel to it) will remain undistorted (Fig. 28). Two-Dimensional Transformations Viewplane in xy plane A += Object ‘ Dop z oe Fig. 28 Assume that the viewplane is the xy plane and that in the required image the foreshortening factor in the z direction is ! and lines parallel to the z-axis make an angle of alpha with the x-axis. Thus a cube of side / would appear in the image as follows (Fig. 29) z ie Fig. 29 In the image the point P is (Icos(alpha).sin(alpha)), but this corresponds to the point -1) in three-dimensions hence the required transformation in matrix form using homogeneous co-ordinates is : x 10 cosa 0 x y 01 sina 0 y wrote 9 OT ts 1 bit 0 0d 1 which is a shear transformation. If! = 1 then itis called a Cavalier Projection and if !=% then it is called a Cabinet projection ive Projections 'n the following it is assumed that the Centre of Projection is at the ‘origin and that the Plane is normal to the z-axis at a distance d from the origin (Pig. 4) This is the tion that holds after view orientation and view mapping.