Image Filtering, Convolution, and Edge Detection

This course presents image derivatives, correlation and convolution, Gaussian filtering, sharpening, and classical methods for detecting image edges. It covers Prewitt, Sobel, Marr-Hildreth, and Canny edge detectors, including zero crossings, separability, and the tradeoff between smoothing and localization.

Syllabus

Intro
Definitions
Examples: Analytic Derivatives
Discrete Derivative: Finite Difference
Derivatives in 2 Dimensions
Derivatives of Images
Correlation & Convolution
Image Noise
Gaussian Filter
2-D Gaussian
Practice with linear filters
Sharpening
Edge detection
Origin of Edges
What is an Edge?
Characterizing edges
Intensity profile
Tradeoff between smoothing and localization
Edge Detectors
Prewitt and Sobel Edge Detector
Prewitt Edge Detector
David Marr
Ellen Hildtreh
Marr Hildreth Edge Detector
Finding Zero Crossings
On the Separability of Gaussian
On the Separability of LOG
LOG Algorithm
Quality of an Edge
John Canny
Canny Edge Detector