About this previous year paper
IMAGE ANALYTICS (KDS-061)
Important Questions & Answers – Section-wise
SECTION A
(Attempt all questions – 2 × 10 = 20 marks)
a. What is Digital Image Processing?
Digital Image Processing is the technique of manipulating and analyzing digital images using computer algorithms in order to enhance image quality, extract useful information, or perform automated interpretation.
b. Applications of Digital Image Processing
Digital image processing is widely used in medical imaging, satellite image analysis, biometric systems, industrial inspection, traffic monitoring, and security surveillance.
c. Differentiate opening and closing
Opening is a morphological operation used to remove small objects and smooth boundaries, while closing fills small holes and connects nearby objects in an image.
d. Hit or Miss Transform
Hit or miss transform is a morphological operation used for shape detection. It identifies specific patterns in a binary image by matching structuring elements.
e. Point detection process
Point detection is a technique used to identify isolated pixels or points in an image that differ significantly from their surrounding pixels.
f. Edge detection algorithm
Edge detection algorithms identify boundaries of objects by detecting sharp intensity changes in an image. Examples include Sobel, Prewitt, and Canny algorithms.
g. Full form of SIFT and definition
SIFT stands for Scale Invariant Feature Transform. It is used to detect and describe local features in images that remain invariant to scale and rotation.
h. Shape numbers
Shape numbers are numerical descriptors derived from boundary directions that uniquely represent object shapes independent of size and orientation.
i. Pattern Classes
Pattern classes are groups of patterns that share similar characteristics or features used in classification tasks.
j. Optimum (Bayes) Statistical Classifier
Bayes classifier minimizes classification error by assigning patterns to classes with the highest posterior probability based on prior probabilities and likelihood functions.
SECTION B
(Attempt any THREE – 10 × 3 = 30 marks)
Q2(a) Digital image and fundamental steps of image processing
A digital image is a two-dimensional array of pixels, where each pixel represents intensity. The fundamental steps include image acquisition, preprocessing, enhancement, segmentation, feature extraction, classification, and interpretation.
Q2(b) Morphological image processing
Morphological processing focuses on shape and structure of objects in an image. Operations like erosion, dilation, opening, and closing modify images based on structuring elements to remove noise and enhance shapes.
Q2(c) Image segmentation
Image segmentation divides an image into meaningful regions based on similarity in intensity, color, or texture. It is essential for object recognition and analysis.
Q2(d) Feature extraction
Feature extraction identifies important characteristics such as edges, corners, texture, and shape. These features reduce data size and improve classification accuracy.
Q2(e) Pattern and prototype matching
Patterns are sets of measurements describing objects. Prototype matching classifies patterns by comparing them with stored reference patterns using distance measures.
SECTION C
Q3(a) Image acquisition
Image acquisition involves capturing images using sensors such as cameras or scanners and converting them into digital form through sampling and quantization.
Q3(b) Smoothing vs sharpening filters
Smoothing filters reduce noise and blur images, while sharpening filters enhance edges and fine details by emphasizing high-frequency components.
Q4(a) Erosion and Dilation
Erosion shrinks objects by removing boundary pixels, while dilation expands objects by adding pixels to boundaries.
Q4(b) Morphological reconstruction
Morphological reconstruction extracts marked regions of an image using repeated dilation constrained by a mask image. It is used for object extraction and noise removal.
Q5(a) Region growing vs region splitting
Region growing starts with seed points and expands regions, while region splitting divides the image into smaller regions until homogeneity is achieved.
Q5(b) Snake sets and Level sets
Snake sets are active contour models used for boundary detection. Level sets represent contours as mathematical functions and handle topological changes efficiently.
Q6(a) Topological and texture descriptors
Topological descriptors describe connectivity and holes in objects, while texture descriptors analyze surface patterns using statistical measures.
Q6(b) Boundary preprocessing and Fourier descriptors
Boundary preprocessing smooths and normalizes boundaries. Fourier descriptors represent shapes using frequency components.
Q7(a) Multilayer neural networks vs deep CNNs
Multilayer feedforward networks use fully connected layers, while deep convolutional neural networks use convolution layers for spatial feature learning and perform better in image tasks.
Q7(b) Neural networks and deep learning
Neural networks mimic biological neurons for pattern recognition. Deep learning uses multiple hidden layers to learn complex features automatically.