Corner Point Extraction in Images Using the SUSAN Algorithm

This program is developed using MATLAB and employs the SUSAN (Smallest Univalue Segment Assimilating Nucleus) algorithm to extract corner points from images. The implementation utilizes circular masking with a 37-pixel template for nucleus comparison, applying adaptive thresholding to detect geometric features based on grayscale similarity. The compressed package includes complete MATLAB source files (.m), sample images, and configuration scripts for immediate deployment. The algorithm functions by analyzing univalue segments around each pixel nucleus, where corners are identified through USAN (Univalue Segment Assimilating Nucleus) area minimization. Key functions include: - susan_corner_detector.m: Main function handling Gaussian smoothing and corner response calculation - adaptive_threshold.m: Dynamically adjusts threshold based on local image statistics - non_maximal_suppression.m: Eliminates duplicate corner candidates This tool enables rapid and precise corner localization, making it valuable for computer vision applications like image registration, 3D reconstruction, and object recognition. Students, researchers, and engineers can extend the codebase through modular integration of custom filters or additional feature detectors. The implementation supports grayscale/color images and includes batch processing capabilities for multiple image files.