Edge detection technique for micro accessory edges having statistical characters

An edge detection method based on macroscopic and microscopic detection is proposed. Firstly, based on fractal theory, regions are divided and edge regions are detected. The second step is to detect the edge point by microcosmic operator. Then the image edge is curved out by LSM (least square method). Experimental results showed that the method can detect the micro accessory edges with statistical characters exactly from being affected by underside edges, background and random edges. Its detecting precision can reach at the pixel-level.