Edge detection by scale multiplication in wavelet domain

摘要: This paper proposes a wavelet based edge detection scheme by scale multiplication. The dyadic transforms at two adjacent scales are multiplied as product function to magnify the structures and suppress noise. Unlike many multiscale techniques that first form maps several then synthesize them together, we determined edges local maxima directly in after an efficient thrsholding. It is shown multiplication achieves better results than either of scales, especially on localization performance. dislocation neighboring also improved when width filter set large smooth Experiments natural images compared with Laplacian Gaussian Canny algorithms.