HEVC帧间预测与CU划分快速算法

发布时间：2018-01-31 00:48

本文关键词： HEVC模式选择帧间预测块划分　出处：《西安电子科技大学》2014年硕士论文　论文类型：学位论文

【摘要】：HEVC(High Efficiency Video Coding)是由两大视频编码标准制定者ITU-T和ISO/IEC联合制定的下一代新标准。随着视频编码质量要求越来越高，H.264已经不能满足人们的需求。但与H.264相比，HEVC的编码效率提升的时候也带来了一些比较棘手的问题，高计算复杂度的增加尤为明显。如何降低HEVC编码算法的复杂度在保证编码的性能的基础上，成为HEVC标准研究的一个重要问题。为了提升编码性能，本文首先介绍了HEVC编码标准中一些基本概念，然后针对其中的帧间模式预测和CU块划分算法进行了深入研究。本文关于这两部分内容分别从PU和CU的角度提出了两种快速算法。第一种先对视频序列的运动特征进行统计，得到合适的实验值，接着分析了前一帧中对应块的大尺寸CU与当前块中小尺寸CU的PU模式的相关性，从中尽可能多的减少PU的冗余。与HM12.0测试模型中的标准算法相比，在全I帧的情况下，在保证PSNR几乎不变的情况下，编码时间减少了54.09%。第二种考虑到视频帧的纹理特性和编码中所采用的量化参数影响最优编码单元(CU)模式的选择，通过这些因素来提前终止CU的划分。实验结果表明与HM-12.0原始算法相比，，在全I帧以及PSNR不变的情况下，编码时间减少了37.42%，编码速度得到显著提升。
[Abstract]:HEVC(High Efficiency Video coding). ITU-T and ISO/IEC are the next generation of new video coding standards developed by the two major video coding standards. With the quality of video coding requirements higher and higher. H.264 has not been able to meet the needs of people, but compared with H. 264, the coding efficiency of HEVC has also brought some thorny problems. The increase of high computational complexity is especially obvious. How to reduce the complexity of HEVC coding algorithm has become an important problem in the research of HEVC standard on the basis of ensuring the performance of coding. In order to improve the coding performance, this paper first introduces some basic concepts in the HEVC coding standard. Then, the inter-frame mode prediction and CU block partitioning algorithm are studied in depth. In this paper, two fast algorithms are proposed from the point of view of pu and CU, respectively. The first one is for video sequence. The movement characteristics were counted. Then the correlation between the large size CU of the corresponding block in the previous frame and the pu mode of the current medium and small size CU of the current block is analyzed. Compared with the standard algorithm in the HM12.0 test model, it can reduce the redundancy of pu as much as possible. In the case of full I frame, the PSNR is almost unchanged. The coding time is reduced by 54.09. The second method takes into account the texture characteristics of the video frame and the quantization parameters used in the coding to influence the selection of the optimal coding unit (CU) mode. The experimental results show that compared with the original HM-12.0 algorithm, the coding time is reduced by 37.42% in the case of full I frame and PSNR invariant. The coding speed was significantly improved.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN919.81

【参考文献】