A New Matching Cost Function for Similar Texture Images Using Segmentation Based Census

It is a challenge to match two corresponding points between images captured from two little different views. Particularly, if some regions in the image have similar textures, finding correspondence will become very difficult. Traditional cost functions have not been competent for this work. To tackle this problem, this paper presents a new cost function using segmentation based census to obtain accurate initial matching. Experimental results show that our method can obtain better performance than conventional ones and preserve clear object border.