基于 3D 相邻区域宏块相关性的多视点视频快速编码算法

针对多视点视频编码的高复杂度，提出一种利用相邻宏块编码信息的相关性进行编码的快速算法。利用多视点视频相邻宏块率失真代价分布的相关性，对当前宏块进行SKIP模式下的提前判决；根据视点间、时间和空间相邻区域运动矢量差值对当前宏块运动类型进行快速分类；针对不同的运动类型调整搜索范围，选择相应尺寸大小的编码模式。通过对具有不同运动特性的多视点视频编码性能测试，结果表明，相对于联合多视点视频模型（joint multi view video model，JMVM）遍历模式，在增加0．75％输出比特率和降低0．04 dB峰值信噪比的情况下，该快速编码算法能平均减少66％的编码时间。

Fast multi view video coding algorithm based on the relativity of 3D adjacent areas macroblock

Aimed at the high complexity of multi view video coding, a fast algorithm based on the correlation of adjacent macro-block''s coding information is proposed in this paper. Using the relativity of adjacent macro-block''s rate distortion cost distribution in inter views of multi view video, current macro-block''s SKIP mode can be prejudge. According to the motion vector difference in the adjoining domains from visual, temporal and spatial perspective, current macro-block''s motion activity can be classified quickly. Search range is adjusted and suitable mode size is chosen in accordance with the different motion activity. Through coding performance testing on multi view video sequences that has respective motion activity characteristics, the simulation results show that this algorithm achieves a reduction of the encoding time by 66% on average, while incurring only 0. 04 dB loss in peak signal-to-noise ratio and 0. 75% increment on the total bit rate compare with JMVM ergodic model.