基于HEVC视频标准的运动估计算法研究

发布时间：2019-06-12 13:29

【摘要】：运动估计技术是视频压缩编码中的核心技术之一,研究设计高效、快速、鲁棒的运动估计算法是视频压缩技术研究的重要课题。在各类运动估计算法中,基于块的运动估计兼顾运动估计精度和复杂度,且计算量小,简单易实现,因此被视频编码国际标准采用。H.264/AVC和H.265/HEVC都是基于块的混合编码结构,基于块的运动估计是其核心算法。视频标准HEVC相对于前一个视频标准H.264压缩效率提高了50%以上,但是计算复杂度也是H.264的三倍。本文旨在研究不降低视频质量并保持HEVC码率条件下,构建能够降低运动估计算法计算复杂度的算法。本论文主要是在分析和研究经典的运动估计算法,以及近年来出现的运动估计优化算法的基础上,改进并实现一种更优的算法。本文详细阐述了视频压缩编码中的运动估计过程,然后详细分析了运动估计过程中涉及到的主要计算,对这些计算的特点进行分析从而进行相应改进,并通过与HEVC中的TZsearch算法进行对比实验来验证这些改进的有效性,最后提出一种改进的基于预测信息的旋转六角模式搜索算法。此算法根据运动矢量间的时空相关性进行运动矢量预测,利用这些预测信息得到最佳的搜索窗口和搜索起点,以及一个合理的终止阈值,在几乎没有损失的情况下,有效地减少了迭代次数。在搜索模式上,提出了一种步长可变的旋转六角模式,兼顾水平与垂直方向上的搜索,并在搜索过程中用精度更高的值替换原来的阈值,做到阈值自适应。本文算法用HEVC参考软件进行测试,测试结果表明本文提出的基于预测信息的旋转六角模式搜索算法相比TZsearch算法运动估计时间减少高达44.7%,而PSNR、码率基本与TZsearch算法相同。通过与FS与TZsearch算法的对比实验,可以得出以下结论:本文算法保持了TZsearch快速搜索算法在搜索精度方面优异的性能,并且大幅降低了运动估计的计算复杂度。
[Abstract]:Motion estimation is one of the core technologies in video compression coding. In that various motion estimation algorithm, the motion estimation accuracy and the complexity are taken into account based on the motion estimation of the block, and the calculation amount is small and is simple and easy to realize, so that the video coding international standard is adopted. The H.264/ AVC and H.265/ HEVC are block-based hybrid coding structures, and the block-based motion estimation is its core algorithm. The video standard HEVC is increased by more than 50% relative to the previous video standard H.264 compression efficiency, but the computational complexity is also three times that of the H.264. The purpose of this paper is to study the algorithm to reduce the computational complexity of the motion estimation algorithm without reducing the video quality and keeping the HEVC code rate. This paper is based on the analysis and research of the classical motion estimation algorithm, and the optimization algorithm of the motion estimation that has occurred in recent years. In this paper, the motion estimation process in the video compression coding is described in detail, the main calculation involved in the motion estimation is analyzed in detail, the characteristics of these calculations are analyzed, and the corresponding improvement is carried out. The effectiveness of these improvements is verified by a comparison experiment with the TZsearch algorithm in HEVC, and an improved rotating hexagonal pattern search algorithm based on the prediction information is proposed. The algorithm performs motion vector prediction according to the space-time correlation between the motion vectors, uses the prediction information to obtain the optimal search window and the search starting point, and a reasonable termination threshold, and effectively reduces the number of iterations without loss. In the search mode, a variable-step-step-step rotation hexagonal mode is proposed, which takes into account the search in the horizontal and vertical directions, and replaces the original threshold with the higher-precision value in the search process, so that the threshold self-adaptation is achieved. The algorithm is tested with the HEVC reference software. The test results show that the motion estimation time of the TZsearch algorithm is up to 44.7% compared with the TZsearch algorithm, and the PSNR and the code rate are basically the same as the TZsearch algorithm. Compared with the FS and TZsearch algorithm, the following conclusions can be drawn: the algorithm maintains the excellent performance of the TZsearch fast search algorithm in the search precision, and greatly reduces the computational complexity of the motion estimation.
【学位授予单位】：湘潭大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN919.81

【相似文献】