Color-compensated Stereo Matching Algorithm for Color Inconsistent Stereo Pair

In conventional algorithms for a stereo matching, corresponding color information between a pair of images is mainly used to find matching points. However, due to various reasons, color information of both images cannot be matched accurately in most practical cases. In this paper, we propose a color-compensated stereo matching algorithm for severely distorted stereo images. Initial corresponding points are extracted by a feature-based matching method from histogram-equalized stereo images. From this information, we estimate relation of color formation of stereo images with a sigmoid function, and compensate for differences. Then, we apply the block-based stereo matching algorithm using the graph cut method. Our method shows good performance when input images are severely distorted.