Digital Image Processing Course

Lesson 1 • Nature of Digital Images

Defines what a digital image is and how continuous scenes are discretized into pixel grids. Anchors all subsequent processing concepts in a concrete spatial model.

Lesson 2 • Histograms and Pixel Statistics

Introduces intensity histograms as diagnostic tools for exposure and contrast analysis. Connects statistical image description to enhancement decisions in later chapters.

Lesson 3 • Color Models and Spaces

Covers RGB, HSV, YCbCr, and grayscale representations and their practical trade-offs. Enables informed color space selection for filtering and segmentation tasks.

Lesson 1 • Color Image Enhancement

Applies point operations in HSV and LAB spaces to enhance color images without hue distortion. Connects grayscale enhancement theory to multichannel images.

Lesson 2 • Adaptive Histogram Equalization

Introduces CLAHE and tile-based local equalization to handle spatially varying contrast. Extends global methods to images with uneven illumination.

Lesson 3 • Histogram Equalization

Explains global histogram equalization using the CDF to redistribute intensity levels. Demonstrates how uniform histograms improve perceptual contrast.

Lesson 1 • Sharpening and Edge Enhancement

Applies Laplacian, unsharp masking, and high-boost filtering to enhance fine detail. Links sharpening to second-derivative operators introduced in this section.

Lesson 2 • Edge Detection Operators

Implements Sobel, Prewitt, and Canny operators to locate intensity discontinuities. Provides foundational edge maps used in segmentation and feature extraction.

Lesson 3 • Morphological Filtering Basics

Introduces erosion, dilation, opening, and closing on binary images as shape-based filters. Bridges spatial filtering to structural image analysis in later chapters.

Lesson 4 • Smoothing and Noise Reduction Filters

Covers box, Gaussian, and median filters for noise suppression with trade-off analysis. Prepares students to select filters based on noise type and edge preservation needs.

Lesson 1 • Homomorphic Filtering

Uses the illumination-reflectance model and log-domain filtering to correct uneven lighting. Extends frequency-domain tools to multiplicative image degradation models.

Lesson 2 • Lowpass and Highpass Filters

Designs ideal, Butterworth, and Gaussian lowpass and highpass filters in the frequency domain. Demonstrates ringing artifacts and smooth roll-off trade-offs.

Lesson 3 • Frequency Domain Filtering Pipeline

Covers the multiply-in-frequency approach: transform, filter, inverse transform, and post-processing. Connects DFT theory to practical filter implementation.

Lesson 4 • Bandpass, Bandreject, and Notch Filters

Constructs bandpass, bandreject, and notch filters to isolate or suppress specific frequency bands. Applies these filters to remove periodic noise patterns.

Lesson 1 • Inverse and Pseudo-Inverse Filtering

Derives the inverse filter and explains its instability due to zero-crossings in the PSF spectrum. Motivates regularized alternatives by demonstrating noise amplification.

Lesson 2 • Noise Estimation and Removal

Estimates noise variance from flat image regions and applies mean, median, and adaptive filters. Connects noise statistics to filter parameter selection.

Lesson 3 • Wiener Filter and Regularization

Derives the Wiener filter using signal-to-noise ratio and introduces Tikhonov regularization. Provides practical tools for restoring blurred images with known noise levels.

Lesson 4 • Blind Deconvolution and PSF Estimation

Covers iterative PSF estimation methods when the blur kernel is unknown. Prepares students for real-world restoration where degradation parameters must be inferred.

Lesson 1 • Active Contours and Level Sets

Introduces snake models and level-set methods for deformable boundary segmentation. Enables precise contour fitting in medical and scientific imaging contexts.

Lesson 2 • Region-Based Segmentation

Implements region growing, region splitting, and merging algorithms based on homogeneity criteria. Extends thresholding to spatially connected region analysis.

Lesson 3 • Watershed and Morphological Segmentation

Uses gradient magnitude images and watershed flooding to delineate object boundaries. Addresses over-segmentation with marker-controlled watershed.

Lesson 4 • Thresholding Methods

Covers global Otsu thresholding, multi-level thresholding, and local adaptive methods. Provides the simplest segmentation baseline for high-contrast images.

Lesson 1 • Local Feature Descriptors

Builds SIFT, SURF, and ORB descriptors that encode gradient orientation histograms around keypoints. Enables robust matching across viewpoint and illumination changes.

Lesson 2 • Scale-Space and Blob Detection

Introduces Gaussian scale-space, LoG, and DoG blob detectors for scale-invariant keypoints. Extends point detection to multi-scale feature localization.

Lesson 3 • Feature Matching and Homography

Applies nearest-neighbor matching, ratio test, and RANSAC to find reliable correspondences. Connects feature description to geometric transformation estimation.

Lesson 4 • Corner and Interest Point Detection

Covers Harris, Shi-Tomasi, and FAST detectors for locating stable keypoints. Establishes repeatable point detection as the first stage of feature pipelines.

Lesson 1 • Object Detection Architectures

Covers anchor-based and anchor-free detectors including YOLO and Faster R-CNN frameworks. Extends classification CNNs to localization and multi-object detection tasks.

Lesson 2 • Convolutional Neural Network Fundamentals

Explains convolution layers, pooling, activation functions, and fully connected layers in CNNs. Connects classical spatial filtering to learned convolutional feature extraction.

Lesson 3 • Training CNNs on Image Data

Covers loss functions, backpropagation, optimizers, and regularization for image classification. Prepares students to train models from scratch and diagnose convergence issues.

Lesson 4 • Deep Image Restoration and Generation

Applies encoder-decoder networks, U-Net, and GANs to denoising, super-resolution, and synthesis. Integrates deep learning with classical restoration concepts from earlier chapters.