Novel scheme for disparity estimation and occlusion detection based on variable line-segment primitive

Disparity estimation and occlusion detection is a crucial issue in stereoscopic and multi-view video processing. But accurate dense disparity fields and reliable occlusions are very difficult to obtain. In this paper, a novel scheme for disparity estimation and occlusion detection is proposed. The scheme selects variable line segment as feature primitive. The uniqueness constraint and ordering constraint are employed to enhance the accuracy of disparity estimation and the reliability of occlusion detection. The theorem of the new scheme is analyzed and experiments are implemented. Experimental results reveal that the proposed algorithm can acquire reliable, sub-precise and dense disparity fields. The detected occlusions are reliable. Comparison results reveal that the proposed scheme is superior to other common algorithms