A fast stereo matching using image segmentation for high quality dense disparity maps

A novel stereo matching algorithm using image segmentation is proposed in this paper. We investigate disparity distribution models (DDMs) on each segment for achieving very high quality results quickly without global optimization. A novel disparity plane fitting strategy based on the investigation is developed for accurately estimating disparity planes. We assign a frontal horizontal disparity plane to a segment as a supplement to the strategy. The experimental results show the outstanding performance of our algorithm both in accuracy and efficiency.