Scale correlation-based edge detection

This paper proposes an effective edge detection scheme based on the scale correlation in the wavelet transform that is equivalent to Canny edge detection. A correlation function is defined as the product of two adjacent wavelet subbands to magnify edges while filtering noise. Unlike many multi-scale techniques that form the edge maps on different scales and then synthesize them into a spatial edge map, our scheme detects edges as local maxima directly in the correlation function. The scheme totally avoids the potentially ill-posed synthesizing operation. The product of detection and localization criteria is higher than that of a single scale, which results in means better edge detection. The dislocation of neighboring edges is also improved. The performance of the presented scheme is illustrated by synthetic and natural images.