新型高效率视频压缩标准HEVC帧内编码算法研究

发布时间：2018-01-18 10:32

本文关键词：新型高效率视频压缩标准HEVC帧内编码算法研究　出处：《华中师范大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着多媒体技术的快速提升,高清晰度数字视频的应用越来越广,视频数据量亦随之不断加大,然而,由于很难在实际应用中传输巨大的视频数据,导致对视频高压缩比技术的迫切要求日益提升。因此,视频编码组(JCT-VC)在考虑到H.264无法满足日益增长需求的基础上,正式制定了新型高效率视频压缩标准HEVC。然而,随着HEVC压缩效率将近一倍的提升,其算法复杂度也随之增加了近2-4倍,这对压缩速度有着巨大的致命损害,因此,在确保HEVC视频质量无明显变化的基础上,如何有效减少HEVC标准的算法复杂度,大幅度地加快视频压缩速度,推进HEVC发展,成为HEVC应用普及化和商业化的问题所在。本文在保证质量可接受情况下,从降低HEVC算法复杂度入手,研究了编码器中较为耗时的地方,根据编码耗时的统计结果,剖析了其编码时间过高的根源,为了加快帧内预测过程和简化模式判决流程,从而提出了两种快速算法：基于SSIM和MPM的快速预测算法和基于候选模式概率的快速模式判断算法。本文所提出算法的主要方向是通过增加一种特性来跳过己有算法,同时,采用多种策略减少候选预测模式的数量,进而在保证性能的前提下,优化其计算复杂度。针对每一个预测块均要使用基于SATD代价函数进行粗略模式选择的问题,本文利用相邻块结构信息相似性SSIM与最佳预测模式相关性的特点以及最有可能模式MPM被确定为最终模式的概率特征,提出了基于SSIM和MPM的快速预测算法,以此来实现不必所有的预测块均要进行粗选模式和避免不必要的率失真优化。针对每个类型PU的候选模式总数之多的问题,本文通过深入研究了不同代价值的模式最终被选为最佳预测模式的概率不尽相同,在此基础上,提出了基于候选模式概率的快速模式判断算法,从而,达到了缩减候选模式集数量的目的,利用五类视频测试序列对本文所提算法和HEVC标准实施对比,实验结果显示,在保持码率无明显变化和主客观质量可接受基础上,该算法可以减少HEVC标准参考代码HM12.1编码时间33.03～46.65%,大大地减少了Intra编码的计算时间,另外,通过RD曲线和重建图像的主观质量对比分析发现,本文所提出的两种算法结合后的压缩性能与HEVC标准原有算法的编码性能几乎不存在差异,说明该算法对于HEVC的实用性和商业性具有实用价值。
[Abstract]:With the rapid improvement of multimedia technology, the application of high-definition digital video is becoming more and more extensive, and the amount of video data is increasing. However, it is difficult to transmit huge video data in practical applications. Therefore, the video coding group JCT-VCC can not meet the increasing demand of H.264. A new efficient video compression standard, HEVC, has been developed. However, with the improvement of HEVC compression efficiency, its algorithm complexity increases 2-4 times. This has a huge fatal damage to the compression speed. Therefore, on the basis of ensuring that the HEVC video quality has no obvious change, how to effectively reduce the algorithm complexity of the HEVC standard. Greatly accelerate the speed of video compression, promote the development of HEVC, become the problem of popularization and commercialization of HEVC applications. Starting with reducing the complexity of HEVC algorithm, this paper studies the time-consuming part of the encoder, and analyzes the root cause of the excessive coding time according to the statistical results of the coding time. In order to speed up the intra prediction process and simplify the mode decision process. Thus, two fast algorithms are proposed:. The fast prediction algorithm based on SSIM and MPM and the fast pattern judging algorithm based on candidate pattern probability. The main direction of the algorithm proposed in this paper is to skip the existing algorithm by adding a characteristic. At the same time, a variety of strategies are used to reduce the number of candidate prediction models, thus ensuring performance. To optimize its computational complexity, every prediction block should use SATD cost function for rough pattern selection. In this paper, we use the characteristics of the similarity of adjacent block structure information SSIM and the correlation between the best prediction pattern and the probability feature that the most likely pattern MPM is determined as the final pattern. A fast prediction algorithm based on SSIM and MPM is proposed. In this way, it is not necessary for all prediction blocks to carry out rough selection mode and avoid unnecessary rate-distortion optimization, aiming at the problem of the total number of candidate modes for each type of pu. In this paper, through in-depth study of the different generations of value of the model is ultimately selected as the best prediction model is not the same probability, on the basis of this, proposed a candidate model based on the probability of fast pattern judgment algorithm, thus. In order to reduce the number of candidate pattern sets, five kinds of video test sequences are used to compare the proposed algorithm with the HEVC standard, and the experimental results show that. On the basis of keeping the code rate unchanged and the subjective and objective quality acceptable, the algorithm can reduce the HM12.1 coding time of HEVC standard reference code by 33.03 ~ 46.65%. The computational time of Intra coding is greatly reduced. In addition, the subjective quality analysis of Rd curve and reconstructed image is found. There is almost no difference between the compression performance of the two algorithms and the coding performance of the original HEVC standard algorithm, which indicates that the proposed algorithm has practical value for the practicality and commercial of HEVC.
【学位授予单位】：华中师范大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN919.81

【参考文献】