HEVC视频编码快速算法的研究

发布时间：2018-03-18 17:04

本文选题：HEVC　切入点：视频编码　出处：《华南理工大学》2014年硕士论文　论文类型：学位论文

【摘要】：新一代视频编码标准HEVC已经成为国际标准，虽然相比上一代视频编码标准H.264/AVC压缩率得到了极大提高，在同等视频质量下码率节省了一半，但高压缩率所带来的高复杂度也影响其广泛的应用，目前HEVC还处在研究发展阶段。对HEVC进行算法优化提高编码的效率是当前HEVC研究中的热点，在保持原有视频质量基本不变的情况下优化HEVC编码器的性能，将可以节省大量的编码时间，从而使其进一步推广应用成为可能。本文针对HEVC编码器进行快速算法的研究，提出一套高效的算法大大降低编码的复杂度，从而更好地应用于HEVC视频点播系统中，节省服务器的编码资源。本文提出的算法包括宏观层和微观层两方面。首先在宏观上，针对HEVC复杂的CU递归划分过程，通过空域和时域上相邻的CU深度信息，对CU划分的递归深度进行预测和裁剪，从而减少部分冗余深度的递归计算，大大减少了CU递归划分的复杂度。在微观上，对每一个递归深度下的帧内或帧间预测过程进行快速终止：对帧内预测模式采用一种类似运动估计三步搜索的快速粗选算法，然后对进行RDOQ计算的预测模式进行快速终止；对帧间运动估计则通过一种基于运动矢量相似性的运动估计快速终止算法，首先计算当前递归深度下编码单元的运动差异值，如果小于一定阈值则提前终止下一递归深度的非2Nx2N模式的运动估计，从而达到快速终止部分冗余运动估计模式的效果。通过实验发现，本文提出的宏观层算法最多节省了约45%的编码时间；微观层算法中帧内预测快速算法最多节省了约24%的编码时间，而运动估计快速终止算法则最多节省了约39%的编码时间。总体上，在损失较小视频质量的情况下，本文提出的两层算法最多节省了66%的编码时间，大大削减了HEVC编码器的计算复杂度，，节省了服务器的编码资源。
[Abstract]:The new generation of video coding standards HEVC has become an international standard, although compared with the previous generation of video coding standard H.264 / AVC compression ratio has been greatly improved, with the same video quality save half the bit rate, However, the high complexity brought by high compression ratio also affects its wide application. At present, HEVC is still in the stage of research and development. The optimization of HEVC algorithm to improve the efficiency of coding is a hot spot in the research of HEVC. Optimizing the performance of HEVC encoder while keeping the original video quality basically unchanged will save a lot of coding time and make it possible to further popularize its application. In this paper, a fast algorithm for HEVC encoder is studied. An efficient algorithm is proposed to greatly reduce the complexity of coding, so that it can be better applied to HEVC VOD system and save the coding resources of the server. The algorithm proposed in this paper includes macro layer and micro layer. Firstly, aiming at the complex CU recursive partition process of HEVC, the recursive depth of CU partition is predicted and clipped by the adjacent depth information of CU in spatial domain and time domain. Thus reducing the recursive computation of partial redundancy depth, greatly reducing the complexity of CU recursive partitioning. The intra-frame or inter-frame prediction process in each recursive depth is terminated quickly: a fast coarse selection algorithm similar to the three-step search of motion estimation is used for intra-frame prediction mode, and then the prediction mode for RDOQ calculation is terminated quickly. For inter-frame motion estimation, a fast motion estimation termination algorithm based on motion vector similarity is used to calculate the motion difference of the coding unit at the current recursive depth. If the threshold is less than a certain threshold, the motion estimation of the next non-2Nx2N pattern with recursive depth is terminated ahead of time, so that the partial redundant motion estimation mode can be terminated quickly. Experimental results show that the proposed macro layer algorithm can save about 45% coding time at most, and the fast intra prediction algorithm can save about 24% coding time at most in the micro layer algorithm. But the fast termination algorithm of motion estimation can save about 39% coding time at most. In the case of less video quality loss, the two-layer algorithm proposed in this paper can save up to 66% coding time. The computational complexity of the HEVC encoder is greatly reduced, and the coding resource of the server is saved.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN919.81

【相似文献】