中文摘要：

为了提高高效率视频编码（HEVC）的编码效率,该文对帧间预测算法进行优化,提出一种基于纹理相似性的快速深度判决算法。随着视频分辨率的提高,视频序列中的空域冗余也随之增加,HEVC通过增加编码块尺寸来提高预测效率,代价就是编码复杂度的急剧增加。通过对视频序列分析可知,图像中的平滑区域与细节区域在相邻帧中存在很大的相关性。该文利用相邻已编码帧的相邻编码块的深度信息,来快速判决当前待编码块的深度信息。对于平滑区域,快速判决待编码最大编码单元的最大深度,以减少小块编码单元的模式判决;对于细节区域,快速判决待编码最大编码单元的最小深度,以减少大块编码单元的模式判决。实验结果表明,与原始的HEVC算法相比,该文所提算法编码比特率平均增加约0.13%以内,PSNR的平均降幅为0.09 d B,算法运行时间平均减少了约50%。

英文摘要：

In this paper, an optimized inter-prediction algorithm is proposed for High Efficiency Video Coding（HEVC） based on video texture similarity. With the increasing of the video resolution, spatial statistical redundancy is increased meanwhile. HEVC improves the coding efficiency by using large coding units, therefore increases the coding complexity significantly. For video sequence, the flat area and texture area are high correlated between adjacent frames. In this paper, previous encoded depth or split depth information is used to determine fast the depth of current encoding coding unit. For flat area, the algorithm predicts the maximum depth to reduce the mode decision for smaller coding units. For texture area, the algorithm can fast determine the minimum depth which can save the mode decision for larger coding units. Experimental result shows that the proposed algorithm can reduce 50% coding complexity compared with original HEVC algorithm. Meanwhile, the average PSNR only decreases 0.09 d B and the average coding rate is increased about 0.13%.