论文部分内容阅读
针对高效率视频编码帧内预测过程中,编码单元四叉树划分算法计算复杂度极高的问题,提出一种基于多重纹理特征的帧内编码单元快速划分算法。首先,使用新自定义的纹理提取方法,提取出多重纹理特征;然后根据多重纹理分布信息,跳过前面不必要的划分和提前终止划分。实验结果表明,与同原始HM12.0相比,快速CU划分算法编码时间平均减少35.09%,码率平均增加1.23%,明显提高了帧内编码效率。
Aiming at the problem of high computational complexity of quadtree coding unit in intra-prediction of high-efficiency video coding, a fast partitioning algorithm of intra-coding unit based on multiple texture features is proposed. First, we use the new self-defined texture extraction method to extract multiple texture features; then, according to the multiple texture distribution information, skip the previous unnecessary segmentation and prematurely terminate the segmentation. Experimental results show that, compared with the original HM12.0, the fast CU partitioning algorithm reduces the average coding time by 35.09% and the code rate by an average of 1.23%, which obviously improves the intra-coding efficiency.