Abstract:

The paper proposes an efficient dense disparity map generation method for rectified image pairs. The scanline on the rectified images are divided intogray-consistent segments, each segment is filled with average gray pixels to construct new scanline pairs. The dense match is computed on the constructed scanlines. The experiments prove that the matching ef- ficiency is improved while fuzziness is reduced.

Key words: rectified image pair;dense disparity map, boundary flow;dynamic programming