PART A - (10 x 2 = 20 Marks)

1. Identify the effects of increasing the sampling frequency on image quality.

 Increasing the sampling frequency generally enhances image quality by
improving detail.
 Reducing artifacts and smoothing gradients.
 But it also comes with increased storage, processing demands.
 Potential diminishing returns depending on the application and viewing
conditions.

2. What is Hue and contrast?

Hue corresponds to a specific wavelength of light within the visible spectrum.

3. List the advantages of Walsh transform.

 Computational Efficiency
 Simplicity
 Binary in Nature
 Orthogonality
 Effective in data compression techniques
 Walsh transform is useful in pattern recognition and feature extraction tasks,
aiding in the analysis and classification of signals and images.

4. Give the conditions for perfect transform.

5. Identify the applications of Homomorphic filtering.

Homomorphic filtering is a powerful tool for enhancing and analyzing images and
signals, particularly when dealing with varying illumination conditions and the need to
emphasize specific features or details.

6. Draw the model of image degradation/restoration process.

7. How is edge detection is used for detecting discontinuities in a digital image?

 Edge detection is a fundamental technique in image processing used to
identify and locate discontinuities or significant changes in intensity in a digital
image.
 These discontinuities often correspond to the boundaries of objects within the
image

8. What is over segmentation of an image?

 Increased Complexity.
 Noise Introduction
 Difficulty in Object Recognition
 Higher Computational Load
Page 1 of 10
9. Differentiate between Lossless and Lossy Compression.
Parameters Lossy Data Compression Lossless Data Compression
Definition The lossy data compression The lossless data compression
technique removes a specified leads to a reduction in file size while
amount and quality of data from the still maintaining the original amount
intended original file (data loss). and quality of the data that it carries.
Restoration This technique cannot restore the This technique can easily restore
of files original amount and quality of the the original quality and amount of
data contained in the original file. data in an available file after
decompression.
Data Quality It compromises the original quality It does not lead to the compromise
of data. of the original data quality.
Data Size The original size of data reduces The original size of data stays intact
after lossy data compression. after lossless data compression.
Algorithms This technique uses algorithms like This technique uses algorithms like
Fractal Compression, DWT Arithmetic Coding, Huffman Coding,
(Discrete Wavelet Transform), DCT LZW (Lempel Ziv – Welch), RLE
(Discrete Cosine Transform), (Run Length Encoding), etc.
Transform Coding, etc.
Uses Mainly MP3 audio, MPEG video, Mainly confidential information and
and JPEG image formats make use sensitive documents make use of
of the technique of lossy data lossless data compression. File
technique. formats like BMP, GIF, RAW, ONG,
etc., also use this technique.

10. Identify the key feature of LZW coding

11. (a) Explain In detail the elements of visual perception. (16)

Elements of visual perception:
 The elementary mechanics of how images are formed and perceived by humans.
 Factors such as how human and electronic imaging devices compare in terms of
resolution and ability to adapt to changes in illumination.
Structure of the human eye:
 The eye is nearly a sphere (with a diameter of about 20 mm) enclosed by three
membranes: the cornea and sclera outer cover the choroid and the retina.
Choroid:
 The choroid lies directly below the sclera
 At its anterior extreme, the choroid is divided into the ciliary body and the iris.
 The central opening of the iris (the pupil) varies in diameter from approximately 2 to 8
mm.
Ciliary body:
 The lens consists of concentric layers of fibrous cells and is suspended by fibers that
attach to the ciliary body.

Page 2 of 10
  It is composed of 60% to 70% water, about 6% fat, and more protein than any other
tissue in the eye.
 Both infrared and ultraviolet light are absorbed by proteins within the lens and, in
excessive amounts, can damage the eye.
 There are two types of receptors: cones and rods.
Cone
 There are between 6 and 7 million
cones in each eye.
 located primarily in the central
portion of the retina, called the
fovea highly sensitive to color.
 Cone vision is called photopic or
bright-light vision.
Rods
 The number of rods is much larger:
Some 75 to 150 million are
distributed over the retina.
 The larger area of distribution, and
the fact that several rods are
connected to a single nerve
ending, reduces the amount of
detail discernible by these
receptors.
 Rods capture an overall image of
the field of view.
 This phenomenon is known as scotopic or dim-light vision.
Distribution of rods and cones in the retina.
 The density of rods and cones for a cross section of the right eye, passing through the
region where the optic nerve emerges from the eye.
 The absence of receptors in this area causes the so-called blind spot.
 Then, their density decreases out to the periphery of the retina. in the eye, is
about 265,000 elements.
 While the ability of humans to integrate intelligence and experience with vision
makes purely quantitative comparisons somewhat superficial.

11. (b) (i) What is a colour model? Explain RGB and /HSI colour models with (10)
necessary diagrams.
 Radiance is the total amount of energy that flows from the light source, and it is usually
measured in watts (W).
 Luminance, measured in lumens (lm), is a measure of the amount of energy that an
observer perceives from a light source.
 Brightness: a subjective (practically unmeasurable) notion that embodies the intensity of
Page 3 of 10
 light.
 RGB is useful for hardware implementations and is serendipitously related to the way in
which the human visual system works
 However, RGB is not a particularly intuitive way in which to describe colours
 Rather when people describe colours they tend to use hue, saturation and brightness
 RGB is great for colour generation, but HSI is great for colour description
Primary colors: Secondary colors :
o Red (R) o Magenta (red plus blue)
o Green (G) o Cyan (green plus blue)
o Blue (B). o Yellow (red plus green

The HSI model uses three measures to describe colours:

 Hue: A colour attribute that describes a pure colour (pure yellow, orange or red)
 Saturation: Gives a measure of how much a
pure colour is diluted with white light
 Intensity: Brightness is nearly impossible to
measure because it is so subjective.
Instead we use intensity. Intensity is the
same achromatic notion that we have seen
in grey level images
Intensity can be extracted from RGB images
 Now consider if we stand this cube on the
black vertex and position the white vertex
directly above it Now the intensity
component of any colour can be determined
by passing a plane perpendicular to the
intenisty axis and containing the colour point.
 The intersection of the plane with the intensity axis gives us the intensity component of
the colour.
 In a similar way we can extract the hue from the RGB colour cube.
 Consider a plane defined by the three points cyan, black and white.

(b) (ii) Show how to convert RGB to HSI model? (06)

Converting from RGB to HIS
 Given a colour as R, G, and B its H, S, and I values are calculated as
follows:
 
  12 R  G R  B 
 if B G 1
 cos  1 
H 
360   if B  G   
R  G  R  BG  B 
2 2



S 1
3
R  G  B
min R,G,B I  13 R  G  B

12.
 (a) List the properties of 2D-DFT and prove any four properties. (16)

12. (b) Define Haar transform. Derive the same for n = 4. What are its (16)
properties?

13. (a) (i) Briefly discuss about Histogram equalization technique. (8)
(a) Show how smoothing spatial filters are used for blurring and noise (8)
(ii) reduction in the image.
Page 4 of 10
13. (b) Explain the following. (16)
i) Inverse filtering
 The inverse filter is a straight forward image-restoration method.
 If know the exact Point-Spread Function (PSF) model in the image-degradation
system and ignore the noise effect, the degraded image can be restored using the
inverse filter approach.
 In practice, the PSF models of the blurred images are usually unknown and the
degraded process is also affected by the noise,
 But the major advantage of inverse-filter-based image restoration is that it is simple.
g(m, n) = f (m, n)∗ h(m, n)+η(m, n)
 Here, f(m, n) represents the original image; h(m,n) represents the degradation
system; η(m,n) is the additive noise term and g(m,n) is the degraded image.
 For simplicity, the coordinates of the image are ignored so that the is given as
 g = Hf+η, the error function is given by η= g− Hf

 Here, wish to obtain minimum and is not constrained in any other way, it can
be termed unconstrained restoration. Equation can be written as

 for , we get

 Taking Fourier transform on both sides of Eq

 The restored image in the spatial domain is obtained by taking the inverse Fourier
transform of Eq

 The advantage of inverse filter is that it requires only the blur point-spread function
as a priori knowledge.
 The inverse filter produces perfect reconstruction in the absence of noise.
 Drawbacks of Inverse Filtering, is that it is not always possible to obtain an inverse.
 Another main drawback of an inverse filter is that an inverse filter will not perform
well in the presence of noise.
 If noise is present in the image, the inverse filter will tend to amplify noise which is
undesirable. In the presence of noise, it is better to go for a Wiener filter.

 Even if know the degradation function, we cannot recover the undegraded image
[the inverse Fourier transform of F(u,v)] exactly because N(u,v) is not known. \
 There is more bad news. If the degradation function has zero or very small values,
then the ratio N(u,v) H(u,v) could easily dominate the term F(u,v).

MINIMUM MEAN SQUARE ERROR (WIENER) FILTERING

Page 5 of 10
 The inverse filtering approach section makes no explicit provision for handling noise.
 An approach that incorporates both the degradation function and statistical
characteristics of noise into the restoration process.
 The method is founded on considering images and noise as random variables, and
the objective is to find an estimate of the uncorrupted image f such that the mean
square error between them is minimized. This error measure is defined as

 where E{i} is the expected value of the argument.

 Based on these assumptions, the minimum of the error function in Eq. (5-80) is given
in the frequency domain by the expression

 where used the fact that the product of a complex quantity with its conjugate is equal
to the magnitude of the complex quantity squared.
 The filter, which consists of the terms inside the brackets, also is commonly referred
to as the minimum mean square error filter or the least square error filter.
 That the Wiener filter does not have the same problem as the inverse filter with zeros
in the degradation function, unless the entire denominator is zero for the same
value(s) of u and v.
 The terms in Eq. (5-81) are as follows:

1. = Fourier transform of the estimate of the undegraded image.

2. = Fourier transform of the degraded image.

3. = degradation transfer function (Fourier transform of the spatial

degradation).

4. = complex conjugate of H(u,v).

5.

6. = power spectrum of the noise

7. = power spectrum of the undegraded image.

 Note that if the noise is zero, then the noise power spectrum vanishes and the
Wiener filter reduces to the inverse filter.
 A number of useful measures are based on the power spectra of noise and of the
undegraded image.

Page 6 of 10
  One of the most important is the signal-to-noise ratio, approximated using frequency
domain quantities such as

 This ratio gives a measure of the level of information-bearing signal power (i.e., of
the original, undegraded image) to the level of noise power.

 The mean square error given in statistical form in Eq can be approximated also in
terms of a summation involving the original and restored images

 In fact, if one considers the restored image to be “signal” and the difference between
 this image and the original to be “noise,” we can define a signal-to-noise ratio in the
 spatial domain as

 The closer f and fˆ are, the larger this ratio will be.
 An approach frequently used when these quantities are not known, or cannot be
estimated, is to approximate by the expression

where K is a specified constant that is added to all terms of .

14. (a) What is the objective of image segmentation? List and explain the (16)
region-based image segmentation technique in detail. Mention two
applications of image segmentation.
Image Segmentation
 Most of the segmentation algorithms, based on one of two basic properties of image
intensity values: discontinuity and similarity.
 In the first category, the approach is to partition an image into regions based on abrupt
changes in intensity, such as edges.
 The second category are based on partitioning an image into regions that are similar
according to a set of predefined criteria. Thresholding, region growing, region splitting
and merging.

Segmentation by Region Growing and by Region Splitting and Merging

 Region growing is a procedure that groups pixels or subregions into larger regions
based on predefined criteria for growth.
 The basic approach is to start with a set of “seed” points, and from these grow regions
by appending to each seed those neighboring pixels that have predefined properties
similar to the seed available.
 The procedure is to compute at every pixel the same set of properties that ultimately will
be used to assign pixels to regions during the growing process.
 If the result of these computations shows clusters of values, the pixels whose properties
place them near the centroid of these clusters can be used as seeds.

Page 7 of 10
  Let: f (x, y) denote an input image;
o S(x,y) denote a seed array containing 1’s at the locations of seed points and
0’s elsewhere; and Q denote a predicate to be applied at each location (x, y).
o Arrays f and S are assumed to be of the same size.
o A basic region-growing algorithm based on 8-connectivity may be stated as
follows.
1. Find all connected components in S(x, y) and reduce each connected component to
one pixel label all such pixels found as 1. All other pixels in S are labeled 0.

2. Form an image such that, at each point (x, y), if the input image

satisfies a given predicate, Q, at those coordinates, and otherwise.

3. Let g be an image formed by appending to each seed point in S all the 1-valued points

in that are 8-connected to that seed point.

4. Label each connected component in g with a different region label. This is the
segmented image obtained by region growing.

REGION SPLITTING AND MERGING

 Grows regions from seed points. An alternative is to subdivide an image initially into a set
of disjoint regions and then merge and/or split the regions in an attempt to satisfy the
conditions of segmentation stated.
 The basics of region splitting and merging.
 Let R represent the entire image region and select a predicate Q.
 This splitting technique has a convenient representation in the form of so-called
quadtrees; that is, trees in which each node has exactly four descendants, as Fig.
shows.
 If only splitting is used, the final partition normally contains adjacent regions with identical
properties.
 This drawback can be remedied by allowing merging as well as splitting.

 That is, two adjacent regions and are merged only if .

 The preceding discussion can be summarized by the following procedure in which, at

any step,

1. Split into four disjoint quadrants any region for which

2. When no further splitting is possible, merge any adjacent regions

for which .
3.Stop when no further merging is possible.

Page 8 of 10
 FIGURE (a) Partitioned image. (b) Corresponding quadtree.
R represents the entire image region.

Image of the Cygnus Loop, want to segment the outer ring of less dense matter.

Characteristics of the region of interest:

 Standard deviation greater than the background (which is near zero) and the central
region (which is smoother).
 Mean value greater than the mean of background and less than the mean of the
central region.
 Predicate

14. (b) Mention different techniques for the representation of shapes in (16)
boundary representation and explain the principle behind them.
Techniques for the representation of shapes in boundary representation are
 Chain Codes
 Signatures
 Boundary San
 Skeletons,
CHAIN CODES
 Chain codes are used to represent a boundary by a connected sequence of straight-
line segments of specified length and direction.
 Typically, a chain code representation is based on 4- or 8-connectivity of the segments.
 The direction of each segment is coded by using a numbering scheme, as in Fig.A

FIGURE
Direction numbers for
(a) 4-directional chain code, and (b) 8-directional chain code.
 If the sampling grid used to obtain a connected digital curve is a uniform quadrilateral
all points of a Freeman code based on are guaranteed to coincide with the points of the
curve.
 The same is true if a digital curve is subsampled using the same type of sampling grid,
as in (b).
 The numerical value of a chain code depends on the starting point.
 Can normalize also for rotation by using the first difference of the chain code instead of
the code itself.

Page 9 of 10
FIGURE (a) Digital boundary with resampling grid superimposed. (b) Result of resampling.
(c) 8-directional chain-coded boundary.

 This difference is obtained by counting the number of direction that separate two
adjacent elements of the code.
 This effect can be reduced by selecting chain elements that are long in proportion to
the distance between pixels in the digitized image, and/or by orienting the resampling
grid along the principal axes of the object to be coded.

SIGNATURES
 A signature is a 1-D functional representation of a 2-D boundary and may be generated
in various ways.
 One of the simplest is to plot the distance from the centroid to the boundary as a
function of angle, as illustrated in Fig.
 The basic idea of using signatures is to reduce the boundary representation to a 1-D
function that presumably is easier to describe than the original 2-D boundary.

 Based on the assumptions of uniformity in scaling with respect to both axes, and that
sampling is taken at equal intervals of u, changes in the size of a shape result in
changes in the amplitude values of the corresponding signature values, e.g., [0,1].
 The resulting signature, although quite different from the r(u) curves in Fig. carries
information about basic shape characteristics.
 For instance, horizontal segments in the curve correspond to straight lines along the
boundary because the tangent angle is constant there.
 A variation of this approach is to use the so-called slope density function as a
signature. This function is a histogram of tangent-angle values.
 Because a histogram is a measure of the concentration of values, the slope density
function responds strongly to sections of the boundary with constant tangent angles
(straight or nearly straight segments) and has deep valleys in sections producing
rapidly varying angles (corners or other sharp inflections).

SKELETONS,

Page 10 of 10
  Like boundaries, skeletons are related to the shape of a region.
 Skeletons can be computed from a boundary by filling the area enclosed by the
boundary with foreground values, and treating the result as a binary region.
 In other words, a skeleton is computed using the coordinates of points in the entire
region, including its boundary.
 The idea is to reduce a region to a tree or graph by computing its skeleton.
 The skeleton of a region is the set of points in the region that are equidistant from the
border of the region.
 The skeleton is obtained using one of two principal approaches:
(1) by successively thinning the region (e.g., using morphological erosion) while
preserving end points and line connectivity (this is called topology-preserving
thinning);
(2) by computing the medial axis of the region via an efficient implementation of the
medial axis transform (MAT).
 The MAT of a region R with border B is as follows: For each point p in R, we find its
closest neighbor in B.
 Instead, the approach is to obtain the skeleton equivalently from the distance
transform, for which numerous efficient algorithms exist.

 Finding approaches for computing the distance transform efficiently has been a topic
of research for many years.
o Numerous approaches exist that can compute the distance transform with
linear time complexity, O(K), for a binary image with K pixels.
o The distance transform of a region of foreground pixels in a background of
zeros is the distance from every pixel to the nearest nonzero valued pixel.
Figure shows a small binary image, and is its distance transform.
o Observe that every 1-valued pixel has a distance transform value of 0
because its closest nonzero valued pixel is itself.
15. (a) Design a coder which a source emits letters from an alphabet (16)
A = {k1, k2, k3, k4, k5} with probabilities P(k1) = p(k3)=0.2, P(k2)=0.4,
P(k4)= P(k5)=0.1, entropy= 2.122 bits /symbol. Find a Huffman code for
this source and the average length of the code and its redundancy.

 The average length for this code is = .4 × 1 + .2 × 2 + .2 × 3 + .1 × 4 + .1 × 4 = 2.2

Page 11 of 10
 bits/symbol.
 A measure of the efficiency of this code is its redundancy—the difference between the
entropy and the average length.
 In this case, the redundancy is 0.078 bits/symbol.
 The redundancy is zero when the probabilities are negative powers of two.

15. (b) Illustrate how image processing is applied in face recognition system (16)
with necessary explanation and diagram.
 Image processing plays a crucial role in
various aspects of a face recognition
system, from preprocessing raw images
to extracting features and matching
faces against a database.
 Here's how image processing is applied
in different stages of a typical face
recognition system:

1. Image Acquisition:
 Camera Calibration: Image processing
techniques may be used to calibrate
cameras to correct for distortions and
variations in image quality.
2. Preprocessing
 Normalization: Adjusting the lighting,
scale, and orientation of the face image
to a standard format to improve consistency and comparability.
 Noise Reduction: Applying filters or algorithms to remove noise from the image,
improving the accuracy of subsequent processing steps.
 Face Detection: Using image processing techniques such as Haar cascades or deep
learning-based methods to detect and locate faces within an image.
3. Feature Extraction:
 Landmark Detection: Identifying key facial points (e.g., eyes, nose, mouth) using
image processing techniques to describe the face's geometry.
4. Descriptor Extraction: Extracting features from the face image, such as texture, shape,
or appearance, Face Matching:
 Template Matching: Comparing the extracted features of the input face against a
database of known faces using similarity measures such as Euclidean distance or
cosine similarity..
5. Decision Making:
 Thresholding: Establishing a threshold for similarity scores to determine whether a
match is found between the input face and the database of known faces.
 Identity Verification: Confirming the identity of the person based on the highest
similarity score above a certain threshold.
Example:
Image Preprocessing: The raw input image is preprocessed to normalize lighting, remove
noise, and detect and align faces.
Feature Extraction: Key features such as facial landmarks and descriptors are extracted
from the preprocessed face image.
Face Matching: The extracted features are compared against a database of known faces
using similarity measures or machine learning models.
Decision Making: Based on the similarity scores or classification results, a decision is made
regarding the identity of the input face (e.g., identification or verification).

Page 12 of 10