基于HEVC的视频编码快速算法研究

发布时间：2018-01-18 07:05

本文关键词：基于HEVC的视频编码快速算法研究　出处：《西安邮电大学》2017年硕士论文　论文类型：学位论文

【摘要】：HEVC(High Efficiency Video Coding)是由视频编码专家组 VCEG(Video Coding Experts Group)和动态图像专家组MPEG(Moving Picture Experts Group)组成的视频编码联合组 JCT-VC(Joint Collaborative Team on Video Coding)制订的新一代视频编码标准。HEVC在上一代视频编码标准的基础上引入更多新的技术使其获得更好的图像质量和更高的压缩比,但也带来了巨大的计算复杂度。因此HEVC标准在保证编码性能的同时最大可能的降低自身计算复杂度,是它是否被广泛应用的根本所在。针对HEVC耗时长、复杂度高的问题,本文分别从帧内编码和帧间编码两个方面对HEVC进行优化,论文的主要工作和创新点如下:1.帧内预测编码是HEVC用来消除空间冗余的关键技术,但是复杂的编码单元划分和帧内预测模式选择给帧内编码带来巨大的计算量。从降低计算复杂度的目的出发对帧内预测模块做改进,首先根据相邻编码单元的时空域相关性,缩小编码单元深度遍历范围,提前中止或跳过不必要的深度,同时在选择当前编码单元的最佳模式时,根据Planar模式是否可为最佳模式,采取不同的方案来减少参与率失真代价计算的预测模式数量。实验结果表明,与HM10.0相比,本文提出的算法在保持编码性能的同时将编码时间减少约30%,加快了帧内编码的速度,具有很好的实用性。2.运动估计是帧间预测中的重要部分,70%的编码时间都产生于此过程。搜索策略和匹配准则直接影响运动估计中寻找最佳匹配块的效率和精度。本文将从搜索策略角度出发,提出基于异构钻石模板的TZSearch快速搜索算法。该算法通过分析TZSearch搜索算法中影响性能的主要因素,结合运动矢量的分布特性,首先使用异构钻石模板进行粗略搜索,并在后续的精细搜索中使用垂直水平六边形模板,使其能够动态适应图像自身特征。实验结果表明,与HM10.0相比,本文提出的算法在保持编码性能的同时将总编码时间减少约17%,加快了搜索速度,缩短了运动估计模块的耗时。
[Abstract]:HEVC(High Efficiency Video coding (VCEG) is an expert group on video coding. Video Coding Experts Group and MPEG(Moving Picture Experts Group. JCT-VC(Joint Collaborative Team on Video coding). Based on the previous generation video coding standard, HEVC introduces more new techniques to achieve better image quality and higher compression ratio. But it also brings huge computational complexity, so the HEVC standard not only guarantees the coding performance, but also reduces the computational complexity of itself as much as possible. To solve the problem of long time consuming and high complexity of HEVC, this paper optimizes HEVC from two aspects: intra coding and inter coding. The main work and innovation of this paper are as follows: 1. Intra prediction coding is a key technology used by HEVC to eliminate spatial redundancy. But the intra coding unit partition and intra prediction mode selection bring a huge amount of computation to the intra coding. In order to reduce the computational complexity, the intra prediction module is improved. Firstly, according to the temporal and spatial correlation of the adjacent coding units, the depth traversal range of the coding units is reduced, and the unnecessary depth is stopped or skipped in advance, and the optimal mode of the current coding unit is selected at the same time. According to whether the Planar model is the best one, different schemes are adopted to reduce the number of prediction models to calculate the cost of participation rate-distortion. The experimental results show that compared with HM10.0. The algorithm proposed in this paper reduces the coding time by about 30 percent while keeping the coding performance. It accelerates the speed of intra coding and has good practicability. 2. Motion estimation is an important part of inter frame prediction. The coding time of 70% comes from this process. Search strategy and matching criterion directly affect the efficiency and accuracy of finding the best matching block in motion estimation. This paper will proceed from the point of search strategy. A fast TZSearch search algorithm based on heterogeneous diamond template is proposed, which combines the distribution of motion vectors with the analysis of the main factors affecting the performance of the TZSearch search algorithm. First, the heterogeneous diamond template is used for rough search, and the vertical horizontal hexagonal template is used in the subsequent fine search, which enables it to dynamically adapt to the image's own characteristics. The experimental results show that. Compared with HM10.0, the proposed algorithm reduces the total coding time by about 17 percent while keeping the coding performance, which speeds up the search speed and shortens the time of motion estimation module.
【学位授予单位】：西安邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN919.81

【相似文献】