联合运动估计的多视点视频视差估计新方法①

摘要

The computation complexity of motion estimation and disparity estimation for multiview video encoding was so huge. In order to resolve the problem, a new method for disparity estimation was proposed in the paper, which was combined with motion estimation. In the time domain, Kalman filtering was established for motion state of each view's macroblock, the motion vector of current macroblock was predicted by Kalman filtering. The thesis analyzed the geometry relationship of motion vector and disparity vector in the space domain based on motion estimation, and calculated the present macroblock's parity vector. Experimental results show that the method proposed in this paper not only saves encoding time substantially but also keeps rate distortion, compared to the full search and fast algorithm.%　　针对多视点视频编码中运动估计和视差估计运算量大的特点，提出了一种联合运动估计的多视点视频视差估计方法。在时域上对每个视点的宏块的运动状态建立 Kalman 滤波器，预测当前宏块的运动矢量。在运动估计基础上结合空域信息，分析了视差矢量和运动矢量的几何关系，计算了当前宏块的视差矢量。实验结果表明，本文方法和全搜索算法及快速算法相比，在大幅节省了编码时间基础上又提高了率失真性能。