Edge detection is a very important step of digital image processing, particularly in the areas of feature detection and feature extraction. People have developed a lot of different methods for edge detection.How to compare the performance of different algorithm? There are three criteria which were proposed by Canny in 1986: good SNR, good localization precision and low-responses. And based on the three criteria we use derivative of Gaussian function as optimal filter for edge detection.This paper dose a research for Canny's edge detection criteria. Based on Canny's continuous criteria, I have proposed one discrete criteria,and proofed that the"minimizing multiple responses to a single edge"criterion can be replaced by an appropriate threshold operation. By combining Gauss linear filter and Canny's edge detection algorithm, I have got an integrated edge detection detector. I have implemented the edge detection algorithm by using C++, and proofed its superiority. |