Register No.

SRM Institute of Science and Technology

College of Engineering and Technology
SET - D
School of Computing
(Common to all branches)
Academic Year: 2023-24 (ODD)

Test: CLA-T1 Date: 20-2-2024

Course Code & Title: 21CSE251T DIGITAL IMAGE PROCESSING Duration: 100 minutes
Year & Sem: II Year / IV Sem Max. Marks: 50

Course Articulation Matrix: (to be placed)

Course
S.No. PO1 PO2 PO3 PO4 PO5 PO6 PO7 PO8 PO9 PO10 PO11 PO12 PSO1 PSO2 PSO3
Outcome

1 CO1 3 2 - - - - - - - - - - - 2 -
2 CO2 3 2 - 1 - - - - - - - - - 2 -
3 CO3 3 - 2 - 2 - - - - 1 - - - 2 -
4 CO4 3 2 - 1 - - - - - - - - - 2 -

5 CO5 3 - 2 1 2 - - - - 1 - - - 2 -

Part – B
(4 x 5 = 20 Marks)
Answer All 4 Questions
21a Discuss at least two key preprocessing techniques, such 5 L2 1 1 1.3.1
as noise reduction or contrast enhancement, and explain
how they contribute to improving the quality of images
before further analysis or interpretation. Provide
examples to illustrate the impact of these preprocessing
techniques on image outcomes.

Keypoints:

Noise Reduction: Eliminates unwanted artifacts, like

electronic interference, improving image clarity.

Contrast Enhancement: Adjusts pixel intensity

distribution for better object differentiation.

Image Quality Contribution: Preprocessing enhances

clarity and reliability for accurate analysis.

Analysis Impact: Reduces noise aids in precise

identification; enhanced contrast improves feature
extraction.
OR
21b Discuss the concept of visual perception and its role in 5 L2 1 1 1.3.1
human-computer interaction. Explain the key
components of visual perception, such as image
formation, Brightness Adaption and Discrimination.

Keypoints:

Visual Perception Overview: Visual perception refers

 to how humans interpret and make sense of visual
information. It plays a crucial role in human-computer
interaction by influencing design and usability.

Image Formation: Image formation involves the process

of capturing, processing, and displaying visual stimuli.
Understanding this process helps in designing interfaces
that effectively convey information to users.

Brightness Adaptation: This component of visual

perception refers to the ability of the human visual
system to adjust sensitivity to different levels of
brightness, ensuring consistent perception across varying
lighting conditions.

Brightness Discrimination: It involves the ability to

distinguish between different levels of brightness in
visual stimuli. Designing interfaces with appropriate
contrast and brightness levels enhances readability and
usability for users.
22a Explain the concept of pixel connectivity in digital image 5 L2 1 1 1.3.1
processing. Define and differentiate between 4-
connectivity and 8-connectivity in the context of pixel
relationships. Provide examples to illustrate how these
connectivity concepts are applied in image analysis and
processing.

Keypoints
Pixel Connectivity Overview: Pixel connectivity refers
to the relationship between neighboring pixels in a digital
image. It determines how pixels are connected or
adjacent to each other, influencing various image
processing tasks.

4-Connectivity: In 4-connectivity, pixels are considered

connected if they share a common edge. This results in
pixels being connected horizontally and vertically but not
diagonally.

8-Connectivity: In 8-connectivity, pixels are considered

connected if they share a common edge or corner. This
allows for connections in all eight directions:
horizontally, vertically, and diagonally.

Application Examples: In tasks like image

segmentation, 4-connectivity is used for simpler, blockier
regions, while 8-connectivity is preferred for more
precise delineation, such as in edge detection or boundary
tracing algorithms.
OR
22b Differentiate between image sampling and quantization 5 L2 1 1 1.3.1
processes. Explain how these two processes contribute to
the overall digital representation of an image.

Keypoints:
Sampling Process: Sampling involves selecting discrete
points from a continuous image to create a digital
representation. It determines the spatial resolution of the
digital image.
 Quantization Process: Quantization assigns digital
values to the sampled points, representing the intensity or
color information of the image. It determines the
dynamic range and precision of the digital representation.

Contribution to Digital Representation: Sampling

determines the spatial fidelity of the image, affecting its
sharpness and detail. Quantization determines the color
or intensity resolution, influencing the accuracy and
fidelity of the image representation.

Overall Impact: Together, sampling and quantization

define the digital image's resolution, color depth, and
overall quality, essential for various image processing
and analysis tasks.
23a Compare and contrasting the RGB (Red, Green, Blue) 5 L3 1 1 1.3.1
and HIS (Hue, Saturation, Intensity) color models which
aid them to better understand color models.

Keypoints:
RGB Color Model: Represents colors using
combinations of red, green, and blue primary colors.Each
pixel is defined by its intensity levels of red, green, and
blue, typically ranging from 0 to 255.

HIS Color Model: Represents colors based on their hue,

saturation, and intensity components.Hue represents the
dominant wavelength of light, saturation represents the
purity of the color, and intensity represents the brightness
of the color.

Comparison: RGB is additive, combining primary

colors to produce a wide range of hues, while HIS is
more perceptually uniform, making it suitable for color
manipulations. RGB is device-dependent, varying across
different displays and devices, while HIS is device-
independent, offering consistent color representation.

Contrast: RGB is more commonly used in digital

displays and photography due to its simplicity and direct
mapping to color channels. HIS is often preferred in
image processing tasks like color correction and
segmentation, as it separates color information in a way
that aligns with human perception.
OR
23b Explain the difference between translation, rotation, 5 L2 1 1 1.3.1
scaling, and shearing 2D transformations. Provide a
mathematical representation for each transformation and
discuss how these transformations can be combined to
achieve complex transformations in a 2D space.

Keypoints:
Translation: Translation moves an object from one
position to another in a 2D space. Mathematically, a
translation is represented as T(dx, dy), where dx and dy
are the horizontal and vertical distances by which the
object is moved.

Rotation: Rotation involves rotating an object around a

 fixed point by a certain angle in a 2D space.
Mathematically, a rotation is represented as R(θ), where
θ is the angle of rotation.

Scaling: Scaling changes the size of an object by

stretching or compressing it along the x and y axes in a
2D space. Mathematically, a scaling transformation is
represented as S(sx, sy), where sx and sy are scaling
factors along the x and y axes.

Shearing: Shearing distorts an object by shifting its parts

along one axis in proportion to their distance from a fixed
line in a 2D space. Mathematically, a shearing
transformation is represented as Sh(shx, shy), where shx
and shy are shearing factors along the x and y axes.
24 a Discuss the mathematical formulations of gray level 5 L2 2 1 1.3.1
transformation types and provide examples of situations
where each type of transformation is more suitable.

Keypoints

Mathematical Formulations: Gray level

transformations adjust the pixel intensities of an image
using mathematical functions. Common transformations
include linear scaling, gamma correction, histogram
equalization, and contrast stretching.

Linear Scaling: Suitable for: Adjusting overall

brightness or contrast in an image, especially when the
relationship between input and output intensities is linear.

Gamma Correction: Suitable for: Compensating for

nonlinear display characteristics, enhancing visibility in
dark or bright areas, and improving image quality in
photography or medical imaging.

Contrast Stretching: Mathematical representation:

Stretching the pixel intensity range to cover the entire
dynamic range of display. Suitable for: Improving the
visual appearance of images by expanding the contrast
between the darkest and brightest pixels, often used in
satellite imagery or surveillance systems.
OR
24 b EVPS company is developing a surveillance system that 5 L2 2 1 1.3.1
monitors activity in public spaces using CCTV cameras.
The company uses spatial domain filtering techniques to
enhance the clarity of surveillance images, reducing
noise and improving overall image quality for better
object detection and tracking. Explain the needed filter
functions for the same.

Keypoints:

Spatial domain filtering: This techniques process

images directly in the spatial domain, altering pixel
values based on their local neighborhood. These filters
are applied to enhance image clarity, reduce noise, and
improve overall quality for better object detection and
tracking in surveillance systems.
 Needed Filter Functions:
Median Filter: Removes impulse noise by replacing
each pixel's value with the median value of its
neighborhood.
Gaussian Filter: Smooths the image by convolving it
with a Gaussian kernel, reducing high-frequency noise
while preserving edges.
Mean Filter: Replaces each pixel's value with the
average value of its neighborhood, effectively reducing
salt-and-pepper noise.
Wiener Filter: Adaptive filter that estimates the local
signal-to-noise ratio to attenuate noise while preserving
image details.

Role in Surveillance Systems: These filter functions

play a crucial role in preprocessing surveillance images
to improve object detection and tracking accuracy. By
reducing noise and enhancing image clarity, the filters
ensure that surveillance systems can accurately detect
and track objects of interest in public spaces.

Impact on Image Quality: Implementing these filter

functions helps EVPS company in developing
surveillance systems with higher-quality images.
Enhanced image quality leads to more reliable
surveillance results, aiding in the identification and
tracking of individuals and objects in public spaces.
Part – C
(1 x 10 = 10 Mark)
25 a Describe how to incorporate 2D DFT properties in image 10 L3 1,2 1 1.3.1
processing pipeline to achieve accurate and efficient
quality image.

Keypoints:

Incorporating 2D DFT properties in an image processing

pipeline:

Understanding 2D DFT:

2D Discrete Fourier Transform (DFT) decomposes an

image into its frequency components, providing insight
into its spatial frequency content.

Utilizing Shift Property: The shift property of 2D DFT

states that shifting an image in the spatial domain
corresponds to a phase shift in the frequency domain.This
property is leveraged for operations like image
translation and alignment.

Leveraging Linearity: Linearity property of 2D DFT

allows for additive operations in the frequency domain,
simplifying complex image manipulations. For example,
adding two images in the spatial domain is equivalent to
adding their corresponding frequency spectra in the
frequency domain.

Exploiting Convolution Property: Convolution in the

 spatial domain corresponds to multiplication in the
frequency domain. This property facilitates efficient
implementation of linear filtering operations like blurring
and sharpening using frequency domain convolution.

Enhancing Efficiency with FFT: Fast Fourier

Transform (FFT) algorithms accelerate the computation
of 2D DFT, reducing computational complexity from
O(n^2) to O(n log n). FFT-based filtering techniques are
employed to efficiently apply frequency domain filters to
images.

Applying Frequency Domain Filters:

Frequency domain filters exploit the spectral
characteristics of images to enhance or suppress specific
frequency components. Examples include low-pass, high-
pass, band-pass, and notch filters for tasks like noise
reduction, edge enhancement, and feature extraction.

Leveraging Frequency Domain Sampling: The

frequency domain provides insight into the spatial
frequency content of an image. Sampling in the
frequency domain allows for selective manipulation of
image details based on their frequency characteristics.

Accounting for Aliasing Effects: Aliasing occurs when

high-frequency components in an image exceed the
Nyquist frequency, resulting in spatial frequency folding.
Anti-aliasing filters in the frequency domain mitigate
aliasing effects by suppressing high frequencies before
downsampling.

Optimizing Filter Design:

Designing frequency domain filters involves trade-offs
between frequency response characteristics like
sharpness, passband ripple, and stopband attenuation.
Optimizing filter parameters ensures desired filtering
effects while minimizing unwanted artifacts.

Integrating into Image Processing Pipeline:

Incorporating 2D DFT properties into the image
processing pipeline enhances the accuracy and efficiency
of image manipulation tasks.
By exploiting frequency domain characteristics, quality
images can be achieved with improved clarity, reduced
noise, and enhanced features.
Or
25 b Apply the mathematical operations involved in simple 10 L3 1,2 1 1.3.1
smoothing filter, such as a 3x3 averaging filter,
Minimum and Maximum to an given image f(x,y).
Provide a step-by-step explanation of the convolution
process.

f(x,y) =
Keypoints:

Here's a breakdown of the mathematical operations

involved in applying a 3x3 averaging filter, minimum
filter, and maximum filter to the given image f(x,
y):

3x3 Averaging Filter:

Kernel:

| 1/9 1/9 1/9 |

| 1/9 1/9 1/9 |
| 1/9 1/9 1/9 |

Step-by-Step Convolution Process:

Place the kernel center at each pixel: Slide the kernel

over the image, ensuring its center aligns with each pixel
you want to process.

Multiply corresponding elements: For each pixel

position, multiply the kernel elements with the
corresponding pixel values in the image and sum the
products.

Sum the products: The sum of these products becomes

the new pixel value at the center of the kernel in the
output image.

Repeat for all pixels: Apply steps 1-3 to all pixels in the
image to generate the filtered output.

Consider the top-left pixel (2) in the image. Here's the

convolution process:

2 * (1/9) + 1 * (1/9) + 1 * (1/9) + 3 * (1/9) + 5 * (1/9) + 2

* (1/9) + 5 * (1/9) + 5 * (1/9) + 2 * (1/9) = 2.111
(rounded to three decimal places)

Output Image:
| 2.111 2.222 2.111 2.222 2.111 |
| 2.222 2.333 2.333 2.333 2.222 |
| 2.333 2.444 2.444 2.444 2.333 |
| 2.222 2.333 2.333 2.333 2.222 |
| 2.111 2.222 2.111 2.222 2.111 |

Minimum Filter: No kernel required. For each pixel,

replace its value with the minimum value in its 3x3
neighborhood.

Output Image:

|1 1 1 1 1|
|1 1 1 1 1|
|1 1 1 1 1|
|1 1 1 1 1|
|1 1 1 1 1|
 Maximum Filter: No kernel required. For each pixel,
replace its value with the maximum value in its 3x3
neighborhood.

Output Image:
|5 5 5 5 5|
|5 5 5 5 5|
|5 5 5 5 5|
|5 5 5 5 5|
|5 5 5 5 5|

Key Points:

Smoothing filters like the averaging filter blur the image,

reducing noise and high-frequency details.
Minimum filters preserve edges and suppress high-
frequency components, potentially highlighting dark
regions.
Maximum filters enhance edges and suppress low-
frequency components, potentially highlighting bright
regions.
The choice of filter depends on the desired image
processing goal (e.g., noise reduction, edge
enhancement).

*Program Indicators are available separately for Computer Science and Engineering in AICTE examination
reforms policy.

Course Outcome (CO) and Bloom’s level (BL) Coverage in Questions

Approved by the Audit Professor/Course Coordinator