Disparity map estimation using image pyramid
Mikołaj Roszkowski · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2013
The task of a short baseline stereo matching algorithm is to calculate the disparity map given two rectified images of one scene. Most algorithms assume that a maximal possible disparity exists and search all disparities in the range from 1 to this maximal disparity. In the case of large images and wide disparity search range this can be very computationally demanding. In this article a simple coarse to fine hierarchical matching method based on the Gaussian pyramid and local stereo matching is investigated. Such an approach allows significant reduction of the number of disparities searched compared to the full search algorithm. Moreover it is shown, that grouping pixels into simple square regions is in most cases sufficient to avoid significant errors that typically appear at disparity map discontinuities when hierarchical schemes are used. Finally, it is presented that in most cases the quality of the disparity map obtained using the investigated algorithm is of comparable quality to a disparity map obtained using full-search local stereo algorithm.