HEVC预测单元分割模式自适应快速选择算法

发布时间：2018-01-11 18:19

本文关键词：HEVC预测单元分割模式自适应快速选择算法　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)是在H.264/AVC之后发展起来的新一代视频编码标准,其目标是在H.264/AVC的基础上,将视频压缩效率提高一倍。为了达到更好的编码性能,HEVC采用比较灵活的编码树划分结构,包括编码单元、预测单元和变换单元,并使用8种帧间预测单元分割模式。HEVC为了获得精确的运动参数,在预测过程中采用基于率失真优化的模式遍历搜索方式,这带来了很高的计算复杂度。本论文正是为了解决这种高计算复杂度问题,在保证编码质量的前提下,降低编码复杂度,加快编码速度。现有的预测单元分割模式快速选择算法有CFM、ESD以及Jong-Hyeok Lee提出的算法,这三个算法都在预测单元分割模式的略过上取得了较大成效,但是视频编码时间仍然有待减少,编码复杂度有待降低。本文首先介绍了HEVC编码标准的关键技术,然后针对帧间预测单元分割模式选择方法进行了深入的研究,然后提出了基于空域相关性的预测单元分割模式自适应快速选择算法。HEVC的时域预测结构分为低时延和随机接入,在这两种结构下,视频帧的编码顺序不同,编码图像和参考图像之间的相对距离不同,因此,本论文把算法分为低时延和随机接入两部分进行详细介绍。本算法的主要思想是,对当前编码单元之前所有时域同质的已编码单元,构建编码单元与其最佳空域参考编码单元之间的预测单元分割模式概率表,根据当前编码单元的最佳空域相邻编码单元的预测单元分割模式,按照所构建的概率表自适应地选择预测单元候选分割模式,略过冗余分割模式率失真代价函数的计算过程。实验结果表明,在低时延时域预测结构下,与CFM算法相比,编码时间减少了约15.31%,BD-PSNR仅减少了0.0186dB;与ESD算法相比,编码时间减少了约22.72%,BD-PSNR仅减少了0.04793dB;在随机接入时域预测结构下,与CFM算法相比,编码时间减少了约16.01%,BD-PSNR仅减少了0.00928dB;与ESD算法相比,编码时间减少了约21.33%,BD-PSNR仅减少了0.03367dB;与Jong-Hyeok Lee的算法相比,时间平均减少了9.04%,而BD-PSNR增大了约0.0256dB。综上所述,本文算法在保证编码质量的前提下,减少了编码时间。
[Abstract]:HEVC(High Efficiency Video coding is a new generation of video coding standards developed after H.264 / AVC. The goal is to double the efficiency of video compression on the basis of H.264 / AVC. In order to achieve better coding performance, HEVC adopts a more flexible coding tree partition structure, including coding units. In order to obtain the accurate motion parameters, the prediction unit and the transform unit adopt the mode traversal search method based on rate-distortion optimization in order to obtain accurate motion parameters. This paper aims to solve the problem of high computational complexity and reduce the complexity of coding on the premise of ensuring the quality of coding. Speed up the coding. The existing fast selection algorithms of prediction unit segmentation mode include CFM / ESD and Jong-Hyeok Lee. These three algorithms have achieved great results in predicting cell segmentation mode, but the video coding time still needs to be reduced. The coding complexity needs to be reduced. Firstly, this paper introduces the key technologies of HEVC coding standard, and then makes a deep research on the selection method of inter-frame prediction unit segmentation mode. Then, the prediction cell segmentation mode adaptive fast selection algorithm based on spatial correlation is proposed. The time domain prediction structure of HEVC is divided into low delay and random access, under these two structures. The coding sequence of video frames is different, and the relative distance between the encoded image and reference image is different. Therefore, this paper divides the algorithm into two parts: low delay and random access. The main idea of this algorithm is. A probability table of partitioned mode between the encoding unit and its optimal spatial reference coding unit is constructed for all the homogeneous coded units in the time domain prior to the current encoding unit. According to the prediction unit partition mode of the best spatial adjacent coding unit of the current coding unit, the candidate segmentation mode of the prediction unit is adaptively selected according to the constructed probability table. The experimental results show that the coding time is about 15.31% less than that of CFM algorithm in low time delay domain prediction structure. BD-PSNR only decreased by 0.0186dB; Compared with the ESD algorithm, the coding time is reduced by about 22.72 and only 0.04793 dB; In the random access time domain prediction structure, compared with the CFM algorithm, the coding time is reduced by about 16.01 and BD-PSNR is only reduced by 0.00928 dB; Compared with the ESD algorithm, the coding time is reduced by 21.33 and only 0.03367dB is reduced by BD-PSNR. Compared with the Jong-Hyeok Lee algorithm, the average time is reduced by 9.04 and the BD-PSNR increases by about 0.0256 dB. The algorithm reduces the coding time on the premise of ensuring coding quality.
【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

【相似文献】