Adaptive threshold edge detection with noise immunity by multi-scale analysis

An adaptive threshold edge detection algorithm based on dyadic wavelet transform is presented in this paper. At first a multi-scale edge response function (MERF) is defined as the multiple scales point-wise products of the dyadic wavelet transform to enhance significant image structures and suppress noise. Thereafter, an adaptive threshold is calculated and imposed on the MERF to identify edges as the local maxima of the MERF gradient map without synthesizing the edge maps at several scales together, which was employed in many multi-scale techniques. Experiments on synthetic benchmark and natural images showed that the proposed adaptive threshold multi-scale edge detection algorithm achieves better detection results than that for a single scale, especially on the localization performance; and edge and noise can be better distinguished by MERF comparing with the Mallat wavelet-based multi-scale algorithm and Canny edge detector.