Pipeline stereo matching in binary images

Abstract

In this paper, we present a short review of shape from stereo reconstruction techniques and propose a stereo matching algorithm based on a pipeline approach. At each stage, current positions of match are estimated from previous computed disparities. Determining the best matching only requires the calculus of correlation values of pixels in a small window centered at candidate positions. The algorithm was succesfully applied to stereo pairs of images, but it can be more advantageously applied to sequences of stereo image frames, due to the Pipeline approach. We also present some visual and numeric results obtained using the algorithm and compare them with those obtained by some usual methods, also implemented.