Robust edge detection in noisy images

We describe a new edge detector based on the robust rank-order (RRO) test which is a useful alternative to Wilcoxon test, using r×rr×r window for detecting edges of all possible orientations in noisy images. Our method is based on testing whether a r×rr×r window is spatially partitioned into two subregions having significant differences in local gray-level value between adjacent pixel neighborhoods of a given pixel, using an edge-height model to extract edges of some sufficient height from images corrupted with noises.Some experiments of statistical edge detectors based on the Wilcoxon test and TT-test, and the well-known Canny edge detector with our RRO detector are carried out on synthetic and real images corrupted by both Gaussian and impulse noise. The results show that the performance of the proposed edge detector appears to be the most robust to variations in noise, performing well in all noise distributions tested.