Comparative Analysis of Edge Detection Algorithms: Susan, LoG, and Prewitt Operators

This article presents a comparative analysis of edge detection algorithms, focusing on the Susan operator, Laplacian of Gaussian (LoG) operator, and Prewitt operator. Edge detection algorithms are fundamental techniques in image processing used to identify boundaries and transitions within digital images. The discussed operators represent commonly employed approaches for edge detection tasks. The Susan operator detects edges by analyzing intensity differences between a central pixel and its surrounding neighborhood pixels within a circular mask. In implementation, this typically involves comparing pixel intensities against a threshold to identify corner and edge points, making it effective for noise-resistant edge detection. The Laplacian of Gaussian (LoG) operator employs second-derivative calculations by first applying Gaussian smoothing to reduce noise, followed by Laplacian filtration to detect zero-crossings that correspond to edge locations. This two-step process can be implemented using convolution with predefined kernels or built-in functions like `fspecial('log')` in MATLAB. The Prewitt operator utilizes first-derivative approximations through horizontal and vertical convolution kernels that measure intensity gradients. Code implementation typically involves separate convolutions for x and y directions using kernels such as [[-1,0,1],[-1,0,1],[-1,0,1]] and its transpose, followed by gradient magnitude calculation. By comparing these edge detection operators' characteristics, performance metrics, and implementation requirements, developers can select the most suitable operator for specific application scenarios, considering factors such as computational efficiency, noise sensitivity, and edge localization accuracy.