Performance analysis of a fuzzy disparity selector for stereo matching of image segments under radiometric variations

Stereo matching algorithms generate disparity maps, which contain the depth information of the environment, from two or more images of a scene taken from different viewpoints. The process of obtaining dense disparity maps is a problem which is still being actively researched. The presence of radiometric differences in the images only further complicates the stereo matching problem. In the present research work, the images are initially split into small patches of pixels, such that pixels in each patch have similar intensities. The authors attempt to study the effect of the parameters, namely, tuning parameter 'α' and the number of segments, while the images are subjected to variations in exposure and illumination. The value 'α' performs the function of a weight signifying the contribution of each data cost, when the two data costs are combined in a linear fashion. Lastly, the results of this methodology are compared with other methods that try to tackle the problem of stereo matching under radiometric variations.