Normalized cross correlation for stereo matching under varying illumination

One of the difficulties in obtaining a dense and accurate disparity map through stereo matching is the presence of radiometric variation in the two images used in this process. It is commonly assumed that in stereo matching, the corresponding pixels have similar pixel values. But the presence of this variation causes the regular methods to produce erroneous results. In this paper, we use the concept of normalized cross correlation for the matching purpose which provides a much better result when compared to intensity based methods like sum of absolute differences and sum of squared differences.