Image Analytics (Elective-III)

Unit I

Fundamentals: Fundamental Steps in Image Processing Systems, Image Acquisition, Sampling and Quantization, Pixel Relationships, Mathematical Tools in Digital Image Processing. Intensity Transformation Functions: Image Negatives, Log Transformations, Power-Law Transformations, Histogram Processing. Color Fundamentals, Spatial Filtering — Smoothing and Sharpening Filters. Practical (Python/MATLAB): intensity transformations, histograms, color image smoothing and sharpening.

Unit II

Morphological Image Processing: Erosion and Dilation, Opening and Closing, Hit-or-Miss Transform, Basic Morphological Algorithms, Morphological Reconstruction, Grayscale Morphology. Practical: morphological operations, reconstruction, grayscale morphology.

Unit III

Colour Image Processing & Thresholding: Colour Models, Pseudo-Colour Processing, Full-Colour Processing, Colour Transformation, Smoothing and Sharpening, Colour-Based Segmentation, Active Contours — Snakes and Level Sets, Noise in Colour Images, Colour Image Compression. Thresholding: Basic Global Thresholding, Otsu's Method, Multiple and Variable Thresholding, Region Growing, Region Splitting and Merging. Practical: Otsu's method, smoothing/sharpening, active contour segmentation.

Unit IV

Feature Extraction: Boundary Preprocessing, Boundary Feature Descriptors — Shape Numbers, Fourier Descriptors, Statistical Moments. Regional Feature Descriptors — Topological and Texture Descriptors, Moment Invariants, Principal Components as Feature Descriptors, Whole-Image Features, Scale-Invariant Feature Transform (SIFT). Practical: boundary descriptors, texture descriptors, SIFT.

Unit V

Image Pattern Classification: Patterns and Pattern Classes, Prototype Matching — Minimum-Distance Classifier, 2-D Correlation Matching, Matching SIFT Features, Structural Prototypes, Optimum (Bayes) Statistical Classifiers, Neural Networks and Deep Learning — Perceptron, Multilayer Feedforward Networks, Deep Convolutional Neural Networks. Practical: minimum-distance, Bayes classification, deep CNN.