Edge Detector Based on Wilcoxon Test in Noisy Images

We describe a new edge detector based on Wilcoxon test for detecting edges in noisy images. The idea is to test whether a n×n window is divided into two subregions having significant differences in gray levels, using an edge-height model which is appropriate for detecting edges of some sufficient height in noisy images. Some experimental results of three statistical edge detectors based on the median test, T test and likelihood ratio test with our Wilcoxon detector on sample images both with and without additive noise are given and performances of these detectors are compared in terms of the well-known quantitative measure.