Global Disparity Estimation in Multi-view Video Coding

【摘要】：Multi-view video coding is one of the most interesting technologies for some multimedia applications such as free-viewpoint video, Geo-media platform, geographic information system of Intergraph and so on. In multi-view video coding, global disparity is one of the major properties of the multi-view video, and disparity estimation is a very important method to reduce the view correlation between different views. In this paper, an efficient, robust, and fast global disparity estimation method is described by using eight-parameter perspective and six-parameter affine models, and phase correlation method for initial translation at the top of pyramid. To get the best disparity estimation, the R-D-optimal combination of global and local disparity re-estimation is given in proposed MVC scheme. Experimental results show the comparison of global perspective (affine) disparity estimation to block matching disparity estimation.