Digital Image Fundamentals

CoE4TN3 • Elements of visual perception

Image Processing • Image sensing and acquisition
• Sampling and quantization
• Relationship between pixels
Chapter 2: Digital Image
Fundamentals

The Human Visual System (HVS) Structure of Human Eye

• Why study the HVS? • Eye: Sphere, diameter of 20 mm
– A true measure of image processing quality is how well • Consists of 3 membranes:
the image appears to the observer. 1. Cornea and sclera
2. Choroid
– The HVS is very complex and is not understood well in a 3. Retina
complete sense. However, many of its properties can be • Cornea: transparent
identified and used to our advantage. • Sclera: opaque, connected to cornea
• Choroid: network of blood vessels
• In front choroid is connected to iris diaphragm
• Iris: contracts or expands to control amount of light
• Pupil: central opening of iris, 2 to 8 mm in diameter

3 4

Structure of Human Eye • Retina contains light receptors: Cones & rods
– Cones:
• Lens: • 6 to 7 million,
– focuses light on retina
• located mainly in central part of retina
– Contains 60% to 70% water (fovea)
– Absorbs 8% of visible light • Sensitive to color,
– High absorption in infrared and ultraviolet (can cause damage to eye) • Can resolve fine details because each
• Retina: the inner most layer, covers the posteriori portion of one is connected to its nerve
eye • Cone vision: photopic or bright-light
• When eye is properly focused, light of an object is imaged on – Rods:
the retina • 75 to 150 million,
• No color vision, responsible for low-
• Light receptors are distributed over the surface of retina light vision,
• Distributed a wide region on the retina
• Rod vision: scotopic or dim-light

5 6

1
 Human Eye Image formation in the eye
• Blind spot: a region of retina without receptors, optic nerves • Lens is flexible
go through this part • Refraction of lens is controlled by its thickness
• Fovea: a circular area of about 1.5 mm in diameter • Thickness is controlled by the tension of muscles connected to
• A comparison between eye (fovea) and a CCD camera: the lens
– Density of cones in fovea: 150,000 /mm 2 • Focus on distance objects: lens is relatively flattened,
– Number of cones: 337,000 refractive power is minimum
– A medium resolution CCD chip has the same number of elements in a
• Focus on near objects: lens is thicker, refractive power is
5mm x 5mm area.
maximum
• Perception takes place by excitation of receptors which
transform radiant energy into electrical impulses that are
decoded by the brain.

7 8

Brightness & Intensity Weber Experiment

• The dynamic range of light intensity to which eye can adapt is • To characterize the intensity discrimination properties of eye
enormous - on the order of 1010- from the scotopic threshold  DI starts at zero and is increased slowly
to the glare limit
• The observer is asked to indicate when the circle on the
• Brightness (intensity perceived by visual system) is a constant background becomes visible (just noticeable
logarithmic function of light intensity.
difference).
• HVS can not operate over the entire range simultaneously. It
accomplishes large variations due to brightness adaptation • The ratio DI/I is called the Weber ratio.
• Procedure is repeated for different values of I.

I+DI

9 10

Weber Experiment Intensity & Brightness

• Small values of DI/I: good discrimination • Relationship between brightness and intensity is not a simple
• Large values of DI/I: poor discrimination function!
• Low levels of illumination: high Weber ratio: poor
discrimination
• In high levels of illumination, discrimination improves.

11 12

2
 Intensity & Brightness Image sensing and acquisition
• Mach Band • If a sensor can be developed that is capable of detecting
effect: energy radiated by a band of the EM spectrum, we can image
Although the events in that band.
shades are • Image is generated by energy of the illumination source
constant,
overshoot and reflected (natural scenes) or transmitted through objects (X-
undershoot are ray)
observed near • A sensor detects the energy and converts it to electrical
the transition signals
boundary. • Sensor should have a material that is responsive to the
particular type of energy being detected.

13 14

Image sensing and acquisition Single sensor

• Three principle sensor arrangements: • Most familiar sensor of this type is photodiode
1. Single imaging sensor • In order to generate a 2-D image using a single sensor, there
2. Line sensor has to be relative displacement in both x and y directions
3. Array sensor between the sensor and the area to be imaged

15 16

Sensor strips Sensor strips

• Sensor elements are arranged in a line • Sensor strips mounted in a ring configuration are used in
• Strip provides imaging in one direction, and motion provides medical and industrial imaging to obtain cross-sectional
imaging in the other direction images of 3-D objects (CAT)
• Used in scanners and airborne imaging • Output of the sensors must be processed by reconstruction
• Airborne imaging: imaging system is mounted on the aircraft algorithms to transform the sensed data into meaningful cross-
which flies at a constant altitude over the area to be imaged sectional images.

17 18

3
 Line scan array Sensor arrays
• This is the arrangement used on digital cameras
• Typical sensor for these cameras is the CCD array (Charge
Coupled Devices)
• Since the sensor is two dimensional a complete image can be
obtained
• Motion is not necessary

19 20

A simple image model Sampling & Quantization

• Computer processing: image f(x,y) must be digitized both
• The amount of light that enters the eye depends spatially and in amplitude
on: • Digitization in spatial coordinates: sampling
1. The amount of source illumination incident on the • Digitization in amplitude: quantization
scene, i(x,y) • Image: [f(i,j)]NxM
2. The amount of illumination reflected by the objects • What should be the values of N, M and the number of gray
in the scene, r(x,y) levels G?
f ( x, y )  i ( x, y )  r ( x, y ) • Normally: N=2n , M=2m , G=2k

(x,y): coordinates
Total absorption: r(x,y)=0
Total reflection: r(x,y)=1

21 22

Sampling & Quantization Sampling & Quantization

• Number of bits required to store image: N x M x k

• The more the values of N,M and G: the better approximation
of a continuous image
• Storage and processing requirements increase as well

23 24

4
 Effects of Reducing Spatial Resolution Effects of Reducing Spatial Resolution

25 26

Effects of Reducing Spatial Resolution Effects of Reducing Gray Levels

Effect of reducing spatial resolution: checkerboard pattern

27 28

Effects of Reducing Gray Levels Effects of Reducing Gray Levels

• Effects of Reducing Gray Levels:
• appearance of fine ridge-like structures in areas of
smooth gray levels
• This effect is called false contouring

29 30

5
 Zooming and Shrinkage Zooming
• Zooming: increasing the resolution (size) of an image • A simple way of zooming which works for increasing the size
• Shrinkage: decreasing the resolution of an image of an image by integer numbers is pixel replication
• Example of zooming: we have an image of 500x500 pixels • Visualize assignment in zooming: the enlarged image is
and we want to enlarge it to 750x750 placed on the original image
• Zooming has two steps: creation of new pixel locations and • Gray level of each pixel in the enlarged image is set to the
the assignment of gray levels to those locations gray-level of its nearest pixel in the original image

31 32

Zooming Shrinkage
• A more sophisticated way of accomplishing gray-level • Shrinkage by an integer number can be done by deleting some
assignment is bilinear interpolation of the rows and columns of the image
• v(x’,y’)=ax’+by’+cx’y’+d • Shrinkage by an noninteger factor can be done as the inverse
• The four coefficients are determined from the four equations of zooming
in four known (four nearest neighbors of the point (x’,y’))

33 34

Zooming Relationship between pixels

0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 1 1 1 1 1
0 0 0 0 0 0 1 0 0 0 0 0
0 0 0 0 0 0 1 0 0 0 0 0
0 0 0 0 0 0 1 0 0 0 0 0
0 0 0 0 0 1 0 0 0 0 0 0
0 0 0 0 1 1 0 0 0 0 0 0
0 0 1 1 0 0 0 0 0 0 0 0
0 1 0 0 0 0 0 0 0 0 0 0
0 1 0 0 0 0 0 0 0 0 0 0
0 0 1 1 0 0 0 0 0 0 0 0
0 0 0 0 1 0 0 0 0 0 0 0
0 0 0 0 0 1 0 0 0 0 0 0
0 0 0 0 0 1 0 0 0 0 0 0
0 0 0 0 0 0 1 0 0 0 0 0
0 0 0 0 0 0 0 1 0 0 0 0
0 0 0 0 0 0 0 0 1 0 0 0
0 0 0 0 0 0 0 0 1 0 0 0
0 0 0 0 0 0 0 0 0 1 0 0
0 0 0 0 0 0 0 0 0 1 0 0
0 0 0 0 0 0 0 0 0 1 0 0

35 36

6
 Relationship between pixels Basic relationships between pixels
• Neighbors • A pixel p at coordinates (x,y) has four horizontal and vertical
• Adjacency neighbors:
• Path N4(P)={(x+1,y), (x-1,y),(x,y+1),(x,y-1)}
• Connectivity • The four diagonal neighbors of P
• Region ND(P)={(x+1,y+1), (x-1,y-1),(x-1,y+1),(x+1,y-1)}
• Boundary • The eight point neighbors of P
• Distance N8(P)=N4(P)U ND(P)

37 38

Adjacency Adjacency
• Two pixels are adjacent if they are neighbors and their gray • 4-adjacency: Two pixels p and q with values from V are 4-
levels are similar adjacent if q is in N4(p)
• V: set of gray levels • 8-adjacency: Two pixels p and q with values from V are 8-
• Similar gray level means that the gray levels of both pixels adjacent if q is in N8(p)
belong to set V • 4-adjacency: broken paths
• Exp: • 8-adjacency: multiple paths
– Binary images: V={1}
– Gray level image: V={32,33, …,63,64}
0 1 1 0 1 1 0 1 1
0 1 0 0 1 0 0 1 0
0 0 1 0 0 1 0 0 1

39 40

Adjacency Path
• m-adjacency: Two pixels p and q with values from V are m- • A path from pixel p with coordinates (x,y) to pixel q with
adjacent if: coordinates (s,t) is a sequence of distinct pixels with
q is in N4(p) or coordinates (x0,y0),(x1,y1),…,(xn,yn) where (x0,y0)=(x,y),
q is in ND(p) and the intersection of N4(p) and N4(q) has no pixels with (xn,yn)=(s,t), and points (xi,yi) and (xi-1,yi-1) are adjacent for
values in V. 1 i  n
• n is the length of the path
q q • We can have 4-, 8-, or m-paths depending on the type of
0 1 1 0 1 1 0 1 1 adjacency specified.
0 1 0 0 1 0 0 1 0
0 0 1 0 0 1 0 0 1

p p

41 42

7
 Connectivity Region
• S: a subset of pixels in an image • R: a subset of pixels in an image
• Two pixels p and q are said to be connected in S if there exists • R is called a region if every pixel in R is connected to any
a path between them consisting entirely of pixels in S other pixel in R
• We can have 4-, 8-, or m-connectivity depending on the type • Boundary (border or contour) of a region: set of pixels in the
of path specified. region that have one or more neighbors that are not in R

0 1 1 0 0 1 1 0
0 1 1 1 0 1 1 1
0 0 1 0 0 0 1 0
0 0 0 0 1 0 0 0

43 44

Distance measures Distance measures

• For pixels p,q, and z with coordinates (x,y), (s,t) and (v,w), • D4 distance
respectively, D is a distance functions if:
D4 ( p, q)  x  s  y  t
D ( p, q )  0 • D8 distance
D ( p, q )  D ( q, p )
D8 ( p, q)  max{ x  s , y  t }
D ( p, z )  D ( p, q )  D ( q, z )

Pixel values D4 distances D8 distances

De  [( x  s) 2  ( y  t ) 2 ]1/ 2
0 1 1 2 1 2 1 1 1
0 1 0 1 0 1 1 0 1
0 0 1 2 1 2 1 1 1

45 46

Distance measures Linear & Non-linear operations

• Dm distance: length of the shortest m-path between two pixels • H: an operator whose inputs and outputs are images
• D4, D8 distance between p and q are independent of the pixels • H is linear if for any two images f and g and any two scalars a
along the path and b
• Dm depends on the value of the pixels between p and q H(af+bg)=aH(F)+bH(g)

0 0 1 0 0 1
1 1 0 0 1 0
1 0 0 1 0 0
Dm=3 Dm=2

47 48

8
 Line scan arrays Line scan arrays
• While register is read, image capturing should stop
• A row of photosites forms the imaging device • Readout speed can be increased using more than one register
• Charges of photosites are transferred to a readout register • A 2-D image is formed by relative motion between the scene and sensor
• Readout register works similar to a shift register
Transport register

Gate
Output
N Photosites 2 1
Output
N Photosites 2 1

Gate
Gate
Readout register
Transport register

49 50

Area scan arrays Color imaging with CCD

• Composed of 2-D array of CCD elements.
• Different methods to read the accumulated charge: • Light is separated into red, green and blue components.
• Color filters or prism can be used to break light
Full frame Interline transfer Frame transfer
• Each component is recorded by CCD

Register Register Register

Dark areas: masked elements of the array (not exposed to light)

51 52