Disparity estimation with disparity field correlation and epipolar geometry constraint for multiview video coding

Disparity estimation plays an important role in multiview video coding to eliminate inter-view redundancy with the cost of high computational complexity. In this paper, a fast disparity estimation has been proposed for low complexity multiview video coding utilizing the disparity field´s correlation and epipolar geometry. The spatial-temporal correlation of disparity field is exploited to select the search center and rough searching range firstly. Then, the searching range is further confined according to the information of epipolar geometry. The experimental results show that 58% reduction in coding time has been achieved compared with the original fast-search algorithm in MVC reference software with negligible degradation in rate-distortion performance.