Enhanced Adaptive Canny Algorithm

This document presents an enhanced adaptive Canny algorithm implementation. The algorithm employs an improved switching median filter for noise reduction, which dynamically applies median filtering only to detected noisy pixels rather than processing the entire image. Following denoising, the image is divided into multiple sub-images using grid-based partitioning. The key innovation lies in the adaptive threshold generation mechanism that analyzes both global edge gradient characteristics and local gradient distributions from each sub-image. This dual-scale analysis enables the algorithm to calculate optimal high and low thresholds dynamically through statistical analysis of gradient magnitudes. The implementation typically involves computing gradient maps using Sobel operators, performing non-maximum suppression, and applying hysteresis thresholding with the dynamically generated thresholds. This approach significantly improves edge detection accuracy while maintaining computational efficiency.