A global matching framework for stereo computation

摘要: This paper presents a new global matching framework for stereo computation. In this framework, the second view is first predicted from reference using depth information. A match measure then defined as similarity function between image and actual image. Stereo computation converted into search problem where goal to find map that maximizes measure. The major advantage of visibility constraint inherently enforced in explores several key components including (1) three color segmentation based representations, (2) an incremental warping algorithm dramatically reduces computational complexity, (3) scene constraints such smoothness constraint. Experimental results different types representations are presented. quality computed maps demonstrated through image-based rendering viewpoints.