多视点视频编码中的运动和视差估计快速算法

Fast Motion and Disparity Estimation of Multi-view Video Coding

针对多视点视频编码复杂度高的问题,提出一种基于分层B帧视点-时间预测结构的运动和视差联合估计快速算法.利用参考帧图像和当前编码图像的运动矢量及视差矢量之间的几何关系,设计可靠的预测矢量作为搜索起始点,并分别在不同方向的参考帧内进行小范围的运动补偿以得到最佳运动和视差矢量.该方法由前一次运动/视差估计得到候选矢量来进行下一次视差/运动估计,只需1次搜索过程就能同时确定最佳运动和视差矢量.实验结果表明,该算法与JMVM全搜索算法相比,能在保持编码质量的同时,节省87.69%的编码时间.

To reduce the complexity of multi-view video coding,a fast joint motion and disparity estimation algorithm based on the view-temporal hierarchical B picture coding structure is proposed.By using the geometry correlation between motion and disparity vectors among neighboring view frames,a reliable starting point is found firstly.An iterative search strategy is then carried out around the starting point to get optimal motion vector and disparity vector in both temporal and view directions.The proposed algorithm takes use of the prediction vector of previous motion or disparity field for the next disparity or motion field.It can get both motion and disparity vectors simultaneously during one estimation process.Experimental results demonstrate that the proposed algorithm can successfully save up to 87.69% computational complexity as compared with JMVM full search algorithm with a negligible quality drop.