Adaptive algorithm of conforming image matching

Abstract

This paper presents an adaptive algorithm of conforming image matching based on the principle of conformity. The algorithm consists of several main stages. At the first stage, we find the corresponding points using a minimum value of conformity as the measure of points’ similarity. We define a conformity function as the sum of all possible combinations of squared differences of pixel intensity values on the fragments that are matched. Then, we perform an adaptive procedure of errors correction considering an intensity gradient distribution. An important feature of the algorithm is the finding of error points using a criterion of maximum value of samples’ conformity for every fragment of the disparity map. The results of experiments on the "Teddy" test images are shown.