Improved Canny-Zernike moment-based subpixel edge detection algorithm

Abstract

In order to solve the lack of edge precision of traditional Canny algorithm and the defects of Gaussian filtering and gradient calculation, we propose an algorithm based on wavelet transform to realize image denoising, which improves the detection accuracy by replacing the original gradient template algorithm with the Scharr operator in four directions, and then uses the reference threshold, combined with the Zernike moment algorithm to perform sub-pixel pinpointing of all possible edge points thus The edge detection is achieved. It is confirmed by many experimental results that the algorithm improves the image edge positioning accuracy and processing efficiency, and has better anti-noise ability than the traditional Canny algorithm and Zernike moment algorithm.