CoE4TN4
Image Processing
Chapter 2: Digital Image
Fundamentals
Digital Image Fundamentals
• Elements of visual perception
• Image sensing and acquisition
• Sampling and quantization
• Relationship between pixels
• Mathematical tools used in image processing
2
The Human Visual System (HVS)
• Why study the HVS?
– A true measure of image processing quality is how well
the image appears to the observer.
– The HVS is very complex and is not understood well in a
complete sense. However, many of its properties can be
identified and used to our advantage.
3
Structure of Human Eye
• Eye: Sphere, diameter of 20 mm
• Consists of 3 membranes:
1. Cornea and sclera
2. Choroid
3. Retina
• Cornea: transparent
• 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
4
Structure of Human Eye
• Lens:
– Focuses light on retina
– Contains 60% to 70% water
– Absorbs 8% of visible light
– High absorption in infrared and ultraviolet (can cause damage to eye)
• Retina: the inner most layer, covers the posteriori portion of
eye
• When eye is properly focused, light of an object is imaged on
the retina
• Light receptors are distributed over the surface of retina
5
• Retina contains light receptors: Cones & rods
– Cones:
• 6 to 7 million,
• located mainly in central part of retina
(fovea)
• Sensitive to color
• Can resolve fine details because each
one is connected to its nerve
• Cone vision: photopic or bright-light
– Rods:
• 75 to 150 million
• No color vision, responsible for low-
light vision
• Distributed a wide region on the
retina
• Rod vision: scotopic or dim-light
6
Human Eye
• Blind spot: a region of retina without receptors, optic nerves
go through this part
• Fovea: a circular area of about 1.5 mm in diameter
• A comparison between eye (fovea) and a CCD camera:
– Density of cones in fovea: 150,000 /mm2
– Number of cones: 337,000
– A medium resolution CCD chip has the same number of elements in a
5mm x 5mm area.
7
Image formation in the eye
• Lens is flexible
• Refraction of lens is controlled by its thickness
• Thickness is controlled by the tension of muscles connected to
the lens
• Focus on distance objects: lens is relatively flattened,
refractive power is minimum
• Focus on near objects: lens is thicker, refractive power is
maximum
• Perception takes place by excitation of receptors which
transform radiant energy into electrical impulses that are
decoded by the brain.
8
Brightness & Intensity
• The dynamic range of light intensity to which eye can adapt is
enormous - on the order of 1010- from the scotopic threshold
to the glare limit
• Brightness (intensity perceived by visual system) is a
logarithmic function of light intensity.
• HVS can not operate over the entire range simultaneously. It
accomplishes large variations due to brightness adaptation
9
Weber Experiment
• To characterize the intensity discrimination properties of eye
• ΔI starts at zero and is increased slowly
• The observer is asked to indicate when the circle on the
constant background becomes visible (just noticeable
difference).
• The ratio ΔI/I is called the Weber ratio.
• Procedure is repeated for different values of I.
10
Weber Experiment
• Small values of ΔI/I: good discrimination
• Large values of ΔI/I: poor discrimination
• Low levels of illumination: high Weber ratio: poor
discrimination
• In high levels of illumination, discrimination improves.
11
Intensity & Brightness
• Relationship between brightness and intensity is not a simple
function!
12
Intensity & Brightness
• Mach Band effect:
Although the shades are
constant, overshoot and
undershoot are observed near
the transition boundary.
13
Intensity & Brightness
• Simultaneous contrast:
perceived brightness does not
depend simply on intensity
14
Image sensing and acquisition
• If a sensor can be developed that is capable of detecting
energy radiated by a band of the EM spectrum, we can image
events in that band.
• Image is generated by energy of the illumination source
reflected (natural scenes) or transmitted through objects (X-
ray)
• A sensor detects the energy and converts it to electrical
signals
• Sensor should have a material that is responsive to the
particular type of energy being detected.
15
Image sensing and acquisition
• Three principle sensor arrangements:
1. Single imaging sensor
2. Line sensor
3. Array sensor
16
Single sensor
• Most familiar sensor of this type is photodiode
• In order to generate a 2-D image using a single sensor, there
has to be relative displacement in both x and y directions
between the sensor and the area to be imaged
17
Sensor strips
• Sensor elements are arranged in a line
• Strip provides imaging in one direction, and motion provides
imaging in the other direction
• Used in scanners and airborne imaging
• Airborne imaging: imaging system is mounted on the aircraft
which flies at a constant altitude over the area to be imaged
18
Sensor strips
• Sensor strips mounted in a ring configuration are used in
medical and industrial imaging to obtain cross-sectional
images of 3-D objects (CAT)
• Output of the sensors must be processed by reconstruction
algorithms to transform the sensed data into meaningful cross-
sectional images.
19
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
20
A simple image model
• The amount of light that enters the eye depends
on:
1. The amount of source illumination incident on the
scene, i(x,y)
2. The amount of illumination reflected by the objects
in the scene, r(x,y)
(x,y): coordinates
Total absorption: r(x,y)=0
Total reflection: r(x,y)=1
21
Sampling & Quantization
• Computer processing: image f(x,y) must be digitized both
spatially and in amplitude
• Digitization in spatial coordinates: sampling
• Digitization in amplitude: quantization
• Image: [f(i,j)]NxM
• What should be the values of N, M and the number of gray
levels G?
• Normally: N=2n , M=2m , G=2k
22
Sampling & Quantization
23
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
24
Spatial and Intensity Resolution
• Spatial resolution: smallest discernable detail in an image
• Expressed in dots (pixels) per unit distance
• In printing: dots per inch (dpi)
– Newspaper: 75 dpi
– Magazines: 133 dpi
• Intensity resolution: smallest discernible change in intensity
level
• Number of intensity levels is usually a power of 2
• 8 and 16 bits are common
25
Effects of Reducing Spatial Resolution
26
Effects of Reducing Gray Levels
27
Effects of Reducing Gray Levels
28
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
Image Interpolation
• Image interpolation: used extensively in tasks such as
zooming, shrinkage, rotation, geometric transforms
• Interpolation: estimate values at unknown locations
• Zooming: increasing the resolution of an image
• Shrinkage: decreasing the resolution of an image
• Example of zooming: we have an image of 500x500 pixels
and we want to enlarge it to 750x750
• Zooming has two steps: creation of new pixel locations and
the assignment of gray levels to those locations
30
Zooming
• A simple way of zooming which works for increasing the size
of an image by integer numbers is pixel replication
• Visualize assignment in zooming: the enlarged image is
placed on the original image
• Gray level of each pixel in the enlarged image is set to the
gray-level of its nearest pixel in the original image
31
Zooming
• A more sophisticated way of accomplishing gray-level
assignment is bilinear interpolation
– v(x,y)=ax+by+cxy+d
• The four coefficients are determined from the four equations
in four known (four nearest neighbors of the point (x,y))
• Next level of complexity: bicubic
32
Shrinkage
• Shrinkage by an integer number can be done by deleting some
of the rows and columns of the image
• Shrinkage by an noninteger factor can be done as the inverse
of zooming
33
Shrinkage and Zooming
34
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
Relationship between pixels
• Neighbors
• Adjacency
• Path
• Connectivity
• Region
• Boundary
• Distance
36
Basic relationships between pixels
• A pixel p at coordinates (x,y) has four horizontal and vertical
neighbors:
N4(P)={(x+1,y), (x-1,y),(x,y+1),(x,y-1)}
• The four diagonal neighbors of P
ND(P)={(x+1,y+1), (x-1,y-1),(x-1,y+1),(x+1,y-1)}
• The eight point neighbors of P
N8(P)=N4(P)U ND(P)
37
Adjacency
• Two pixels are adjacent if they are neighbors and their gray
levels are similar
• V: set of gray levels
• Similar gray level means that the gray levels of both pixels
belong to set V
• Exp:
– Binary images: V={1}
– Gray level image: V={32,33, …,63,64}
38
Adjacency
• 4-adjacency: Two pixels p and q with values from V are 4-
adjacent if q is in N4(p)
• 8-adjacency: Two pixels p and q with values from V are 8-
adjacent if q is in N8(p)
• 4-adjacency: broken paths
• 8-adjacency: multiple paths
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
Adjacency
• m-adjacency: Two pixels p and q with values from V are m-
adjacent if:
q is in N4(p) or
q is in ND(p) and the intersection of N4(p) and N4(q) has no pixels with
values in V.
q
q
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
p p
40
Path
• A path from pixel p with coordinates (x,y) to pixel q with
coordinates (s,t) is a sequence of distinct pixels with
coordinates (x0,y0),(x1,y1),…,(xn,yn) where (x0,y0)=(x,y),
(xn,yn)=(s,t), and points (xi,yi) and (xi-1,yi-1) are adjacent for
• n is the length of the path
• We can have 4-, 8-, or m-paths depending on the type of
adjacency specified.
41
Connectivity
• S: a subset of pixels in an image
• Two pixels p and q are said to be connected in S if there exists
a path between them consisting entirely of pixels in S
• We can have 4-, 8-, or m-connectivity depending on the type
of path specified.
42
Region
• R: a subset of pixels in an image
• R is called a region if every pixel in R is connected to any
other pixel in R
• Boundary (border or contour) of a region: set of pixels in the
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
Distance measures
• For pixels p,q, and z with coordinates (x,y), (s,t) and (v,w),
respectively, D is a distance functions if:
44
Distance measures
• D4 distance
• D8 distance
Pixel values D4 distances D8 distances
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
Distance measures
• Dm distance: length of the shortest m-path between two pixels
• D4, D8 distance between p and q are independent of the pixels
along the path
• Dm depends on the value of the pixels between p and q
0 0 1 0 0 1
1 1 0 0 1 0
1 0 0 1 0 0
Dm=3 Dm=2
46
Math tools in image processing
• Linear operators
• Arithmetic operations
• Logic operations
• Spatial/Geometric operations
• Vector and matrix operations
• Image transforms
47
Linear & Non-linear operations
• H: an operator whose inputs and outputs are images
• H is linear if for any two images f and g and any two scalars a
and b
H(af+bg)=aH(f)+bH(g)
48
Arithmetic Operations
• Arithmetic operation between two images is performed
between corresponding pixel pairs
• Example: Averaging of noisy images for noise reduction
49
Logical Operations
• Logical operations are typically applied to binary images.
• We can think of foreground being 1 and background 0.
50
Geometric Transformations
• Geometrical transformations: modify the spatial relationships
between pixels in an image
• Geometrical transformations: called rubber-sheet
transformations
– They may be viewed as printing an image on a sheet of rubber and
then stretching this sheet according to some predefined rules
51
Geometric Transformations
• Geometrical transformation consists of two basic operations:
1. Spatial transformation: defines the rearrangement of pixels on the
image plane
2. Gray level interpolation: deals with the assignment of gray levels to
pixels in the spatially transformed image
52
Spatial Transformations
• Image f with pixels coordinates (v,w) has undergone
geometric transformation to produce an image g with
coordinates (x,y)
• Example: (x,y)=T{(v,w)}=(v/2, w/2)
– Shrinking of the size of f(x,y) by one-half in both directions.
53
Spatial Transformations
• Affine transform
• Can scale, rotate, translate or sheer depending of the values
chosen for elements of T
54
Spatial Transformations
55
Image Registration
• Image registration: to align two or more mages of the same
scene
• The images are available transformation is unknown
• Input image: image the we wish to transform
• Reference image: image against which we want to register
the input image
56
Image Registration
• Solution: spatial relocation is formulated by the use of
tiepoints.
• Tiepoints: a set of pixels whose locations in input and
reference images are known
• Suppose the geometrical transformation process within the
region is modeled by a pair of bilinear equations:
– x=c1v+c2w+c3vw+c4
– y=c5v+c6w+c7vw+c8
– 8 known tiepoints, 8 unknown ci
57
Gray-level Interpolation
• Depending on the values of ci x and/or y can be noninteger
for integer values of (v,w)
– x=c1v+c2w+c3vw+c4
– y=c5v+c6w+c7vw+c8
• We need inferring gray-level values at noninteger locations
(gray-level interpolation)
58
Image Registration
59