Image occlusions as a fuzzy reasoning problem

Abstract

The analysis of images which include partially occluded objects must necessarily deal with incomplete information. Fuzzy reasoning is a natural tool for such an application. In this paper, we formulate a method using fuzzy reasoning to recognize various instances of occlusion. The method makes use of image grey levels, triple point incident angles, and curvature of completed objects to evaluate candidate configurations. Furthermore, we implement this method and evaluate it on real test images.<>