采用率失真与模式特征的多视点视频编码快速模式选择

Fast Mode Decision for Multiview Video Coding using Rate-distortion and Mode Characteristic

  • 摘要: 现有的多视点视频编码使用了分层B帧(Hierarchical B Picture, HBP)的预测结构,其帧内预测、帧间预测以及视点间预测的模式选择给多视点视频编码带来了庞大的计算复杂度。针对这一问题,我们在分析了JMVC模式分布比例的基础上,提出了一个快速帧间模式选择的算法。这种算法利用率失真代价和预测模式特征之间的关系来及时判定最优模式:如果上一尺寸预测模式的率失真代价小于当前尺寸预测模式的率失真代价则认为上一预测模式为最优模式,跳过检查其他更小尺寸的预测模式;反之,如果上一尺寸的预测模式的率失真代价大于当前尺寸的预测模式的率失真代价,则继续检查其他更小的尺寸。这样,通过提前终止一些不必要的模式选择过程,多视点视频编码的计算量得到大幅的降低。实验结果表明:所提算法能在保持JMVC中全搜索算法的编码效率同时,使计算复杂度减少了81.66%。

     

    Abstract: The hierarchical B picture prediction structure adopted by multiview video coding scheme (JMVC) introduces extremely high computational complexity. To speed up the encoding process, the exhaustive mode decision is analyzed first, and then a fast inter frame mode decision algorithm for multiview video coding is proposed in this paper. In our approach, the prediction modes are firstly classified according to their size, and the rate distortion costs are calculated from the largest to the smallest. Based on the correlation of rate distortion cost and the size of prediction mode, the mode decision process can be earlier terminated. If the rate distortion cost of the last (larger size) prediction mode is lower than that of the current (smaller size) mode, skip checking the rest prediction modes and take the last one as the optimal mode. On the contrary, if the rate distortion cost of the last (larger size) prediction mode is higher than that of the current (smaller size) one, continue the checking process. Experimental results have shown that the proposed algorithm can reduce the computational load by 81.66%, while keeping almost the same coding efficiency compared to the full search mode decision of JMVC.

     

/

返回文章
返回