A Comparative Study and Evaluation of Stereo Matching Costs for Radiometric Differences

Stereo correspondence algorithm depends on matching costs in order to compute the similarity of the image locations. The insensitivity of various cost functions for stereo algorithm is evaluated with respect to radiometric variations of the given input image. We consider both the window based variants as well as pixel based variants like the absolute difference, the sampling insensitivity absolute difference, and normalized cross correlation. We measure the performance in the presence of global intensity changes, local intensity changes. Using stereo data sets with ground-truth disparities taken under controlled changes of exposure and lighting, we evaluate the cost with a local and a global stereo method.