A NEW STATISTICAL METHOD FOR EDGE DETECTION ON TEXTURED AND CLUTTERED IMAGES

摘要

A novel method of edge detection using a statistical filter is presented here. The approach uses the application of a set of square masks to scan the image. Each mask is divided into two regions at an angle intersecting the central pixel. The statistical features of these two regions are calculated using a series of tests such as T-Tests and F-Tests. The tests are repeated for each angle to ascertain the likely edge direction. The edge detector is optimised for performance on images that are typically textured or cluttered. The filter has been applied to histological and cluttered concealed-weapons images.