Digital Image Processing

Learn histogram equalization to enhance image contrast by redistributing pixel intensities via the CDF, improving visibility in grayscale and color images, with adaptive variants like AHE and CLAHE.

Presents a step-by-step numerical example of histogram equalization for a 4x4 image, showing histogram, probability distribution, and CDF to map gray levels.

Explore how spatial relationships among pixels define image structure through 4-neighborhood and 8-neighborhood connections. Apply concepts to edge detection, segmentation, and connectivity using Euclidean, city block, and chessboard distances.

Sharpening spatial filters enhance image details and edges by preserving high-frequency components, using gradient operators—Roberts, Prewitt, Sobel, and Laplacian, along with unsharp masking and high-boost filtering.

Compare image enhancement and image restoration, focusing on their goals, methods, and mathematical foundations, where enhancement is subjective and restoration aims to recover the original image.

Identify sources of degradation in digital image processing: sensor noise, motion blur, atmospheric turbulence, and sampling and quantization errors, and model them to guide restoration with inverse or VNR filtering.

Learn how the noise probability density function models image noise and guides restoration by matching histograms to PDFs for Gaussian, salt-and-pepper, and other noise types, including the VNR filter.

Explore mean-based filters for image restoration in the spatial domain, including arithmetic, geometric, harmonic, and contra-harmonic mean filters, their strengths against additive noise, and when to apply them.

Explore order statistics filters for image restoration, focusing on impulse noise and edge preservation, and compare median, alpha trimmed mean, and min/max/midpoint filters with mean filters.

GLCM analyzes texture by counting how often pixel pairs occur at a given distance and direction, then normalizes to produce features such as contrast, energy, homogeneity, and correlation.

Explore local binary pattern, a simple yet powerful texture descriptor that encodes a 3x3 neighborhood into an 8-bit code and builds a texture histogram for image analysis, including face recognition.

Explore lossy compression techniques that reduce image sizes using transform coding and quantization, preserving quality in JPEG and other standards, revealing artifacts and trade-offs.

Discover how lossless compression techniques reduce image storage and bandwidth by removing redundancy while ensuring perfect reconstruction, through methods like run length encoding, Huffman coding, and LZW.

Learn how principal component analysis reduces data dimensionality by rotating into orthogonal components that maximize variance, using standardized data, covariance matrix, and eigenvectors to project into a compact, informative representation.