Features Stereo Matching Based on Fuzzy Logic

This paper presents an entire scheme for the estimation of a sparse disparity map from stereo pair images. In contrary to a dense disparity map, for which disparity values are calculated for each pixel of image, the sparse disparity map is determined only for some distinguished set of pixels from an image. The pixels belong to this set are called features. Therefore, in the first stage the algorithm for determining sparse disparity map has to be able to detect and specify some pixels as the feature pixels. In this article a method for specifying features is introduced. The core of the presented method relies on a fuzzy edges detector. The algorithm of fuzzy edges detection, as well as a manner enabling determination of the features’ set, are introduced. The disparity is calculated at the fuzzy domain based on a similarity measure which is the correlation of fuzzy sets. The results of the obtained disparity maps for some benchmark stereo pairs, and the comparison with the well known Marr-Poggio-Grimson algorithm designed for sparse disparity map estimation are presented. Keywords— feature points, stereo matching, disparity map, edges detection, fuzzy logic