【作者】 郑远；

【导师】 常永宇；

【作者基本信息】 北京邮电大学 ， 通信与信息系统， 2014， 硕士

【摘要】 随着互联网络以及移动互联网络的飞速发展,视频业务成为网络数据业务的一项热门应用,对视频质量评价的研究具有一定实用性。本论文首先介绍了现有的视频质量评价方法。数字视频质量评价方法可以分为两大类：主观质量评价和客观质量评价。直接让观测者对视频质量做出感知判断的主观质量评价算法是最为可靠的评估方法,但该方法实现起来比较复杂且受到诸多因素影响。客观视频质量评价又分为全参考(FR),部分参考(RR)和无参考(NR)。然后,本文介绍了高效率视频编码HEVC的特点和参考软件用法。本文提出了一种基于HEVC编码的无参考PSNR算法并实现。本文提出的方法采用DCT系数拉普拉斯分布模型,针对HEVC编码采用4种不同尺寸变换块的特点,分类计算不同变换块的均方误差,最终得到变换块像素尺寸加权的PSNR估计值,并编程实现了算法。测试结果表明,本文提出的算法求得的无参考PSNR估计值和全参考PSNR计算值有着较好的相关性和一致性,尤其是应用于帧内编码块时性能较好。更多还原

【Abstract】 With the rapid development of Internet and the mobile internet, video service becomes an important part of the network data services. The evaluation of the quality of the video is practical.In this paper, the summary of video quality evaluation is reviewed. Digital video quality assessment methods can be divided into two categories:subjective and objective quality evaluation of quality assessment. The method that observers make directly subjective assessment of video quality is the most reliable approach, but it is too complex to implement and subject to many factors. Objective video quality assessment is divided into full-reference (FR), reduced-reference (RR) and no-reference (NR). Then, this paper introduces HEVC and its reference software. This paper focuses on the no-reference PSNR algorithm based on HEVC coded sequence. The proposed method adopts Laplace distribution model for DCT coefficients, finally obtaining coding blocks’ pixel-number weighted estimated PSNR value, which calculates coding blocks’ PSNR respectively based on four different sizes of transform blocks that HEVC adopts. Experiment results show that the proposed algorithm performs well, estimated PSNR value has a good correlation and consistency with the truth-value, especially in intra-coded blocks.更多还原