Study Of The View Synthesis Techonology For Multiview Video Coding

As dramatical development of wireless technique and IP networks, video communication is becoming one of the major services of multimedia communication. FTV(Free television) and 3DTV(3D television ) is available or will be available in the near future. Finally MVC ( multi-view video coding ) is described as a key component for 3DTV and FTV systems, which was very important to be researched and will be widely applied in various areas.However, in particular, there are still issues to be resolved. For example, the block-based disparity estimation is inaccurate and the pixel-based disparity estimation is inefficient. How to estimate disparity both accurately and efficiently are hotpots which are researched recently. Besides, because of the occluded regions, there may be holes in the synthetic pictures. It is very important to solve these problems.In this paper, we have proposed a novel self-adaptive block size disparity estimation algorithm, which choose the block that obtains the minimum matching cost, to estimate disparities. Several suitable constraints can be applied in disparity estimation in order to enhance efficiency and accuracy. One of constrains is the uniqueness constraint, which implies each image point may be assigned at most one disparity value. We use the left neighborhood, labeled median filter and improved median filter to correct the disparity error respectively.Other most often applied is smoothness constraint, which describes disparity values vary smoothly in neighboring regions. The smooth term need to find the boundary points first, then using the disparity of the neighboring pixel(excluding the boundary points) to smooth every point. The occlude areas of view may cause holes in synthesis pictures, we deals with the holes as follows: judge and choose which view, left or right view, can be used to filling the holes. Then mapping the corresponding pixel and disparity information to the holes so as to fill them.Experimental results demonstrate that the proposed self-adaptive block match approach can improve the operation speed of the disparity estimation, with little change of image quality. Disparity correction and smooth algorithm can improve the quality of the disparity map. Besides, our hole-filling method can also obtain better results.