An Iterative Approach for Disparity Estimation in Multiview Video Applications

Disparity is an important feature of multiview video applications, which can be used in coding techniques, synthesis of views, among others. The disparity estimation, however, is very complex and requires high processing capabilities to produce precise results. This paper presents an estimation method that uses segmentation and disparity information in an iterative cycle. In each cycle the frames are segmented and applied in a disparity estimation process. Error analysis is also made, correcting inaccurate values. The obtained map is then used to improve the segmentation, restarting the cycle. Experimental results indicate that the proposed method works, diminishing up to 50% the mean error presented in the maps.