A Robust Edge Detection Algorithm Based on CR-DSmT

Abstract

By employing the single scale edge detection operators, a small difference between the real edges and the measurement of noise spots for detecting from the reality leads to the problems that the small features may be doped with the noise, and that the efficiency of detecting edge is low. As to the problems, the robust edge detection algorithm based on CR-DSmT is presented. First of all, the multi-scale edge detections of the image are detected and extracted. Then the measures are fused according to the evidence reasoning. Basic Belief Assignment (BBA) construction technique of bi-directional exponential edge detection is offered, which is employed to adjust the belief assignment differences between the real edges and noise by the significant coefficient- , to fuse by the CR-DSmT combination technique, and to determine the final edge pixel by the bi-threshold technique accordingly. Simulation experiments show that it suppresses noise and maintain the detail information in the process of detecting edge.