(SEM VI) THEORY EXAMINATION 2024-25 IMAGE ANALYTICS

IMAGE ANALYTICS (BCDS061)

B.Tech Semester VI – Exam-Ready Notes

SECTION A

(Attempt all | 2 × 7 = 14 marks)
Write 2–3 crisp lines for each answer.

(a) Characteristics of Histogram in an Image

An image histogram shows the distribution of pixel intensity values.
Key characteristics include contrast, brightness, dynamic range, and peaks representing dominant gray levels.

(b) Different Kinds of Filters

Spatial filters (averaging, median, Gaussian)     Frequency domain filters (low-pass, high-pass)

Linear and non-linear filters

(c) Hit-or-Miss Transform

Hit-or-Miss transform is used to detect specific shapes or patterns in binary images by matching a structuring element.

(d) Basic Morphological Algorithms & Applications   Erosion – removes noise and small objects

Dilation – fills gaps and connects regions
Applications include shape analysis and object detection.

(e) Pseudo Color Image Processing

Pseudo color processing assigns artificial colors to grayscale images to enhance visual interpretation of features.

(f) Preprocessing Steps for Boundary Detection

Noise removal (smoothing)                     Image enhancement (contrast improvement)

(g) Key Components of CNN

CNN consists of convolution layers, pooling layers, activation functions, fully connected layers, and output layer.

SECTION B

(Attempt any THREE | 7 × 3 = 21 marks)

(a) Morphological Reconstruction

Morphological reconstruction extracts image components using a marker image and mask image.
Steps: marker selection → iterative dilation → constraint by mask.
Applications: noise removal, object extraction, shape restoration.

(b) Image Smoothing vs Image Sharpening

Smoothing: reduces noise and small variations using filters like averaging or Gaussian.

Sharpening: enhances edges and fine details using high-pass filters.
In color images, smoothing reduces color noise while sharpening highlights boundaries.

(c) Advantages of Deep Learning over Traditional Classification

Automatic feature extraction                      Higher accuracy for complex images

Robust to noise and variations                   Better performance on large datasets

(d) Purpose of Feature Extraction

Feature extraction reduces data dimensionality while retaining important information, making classification and recognition faster and more accurate.

(e) Smoothing Spatial Filters

Averaging filter: reduces noise by replacing pixel with neighborhood mean.

Gaussian filter: smooths image while preserving edges better.
Used in preprocessing and noise reduction.

SECTION C

(Attempt any ONE | 7 marks)

(a) Image Segmentation Techniques

Region Growing: groups pixels based on similarity starting from seed points.

Region Splitting & Merging: divides image into regions and merges similar ones.
Comparison: region growing is simpler, splitting-merging is more systematic but computationally expensive.

OR

(b) Scale-Invariant Feature Transform (SIFT)

Steps:

Scale-space extrema detection                              Keypoint localization

Orientation assignment                                          Keypoint descriptor generation
SIFT is robust to scale, rotation, illumination, and noise.