HEVC屏幕内容的图像恢复的研究

发布时间：2018-04-26 22:06

本文选题：HEVC + 屏幕内容　；参考：《北方工业大学》2017年硕士论文

【摘要】：屏幕内容图像或视频广泛应用于视频会议、远程桌面控制、无线显示等设备中。与照相机捕捉的自然内容不同,屏幕内容通常包含大量的文本、图表以及人造内容,传统的视频编码工具不能很有效地压缩。为此,联合视频编码工作组(Joint Collaborative Team on Video Coding,JCT-VC)致力于开发高效视频编码(High-efficiency video coding,HEVC)标准的屏幕内容扩展工具,例如帧内块拷贝模式、调色板模式、自适应颜色域变换等,并在2016年2月发布了HEVC屏幕内容扩展标准(Screen content coding,SCC)。尽管SCC能够提高屏幕内容编码效率,但是在某些条件下,重建的屏幕内容图像或视频依然存在大量噪声和畸变,降低了重建图像或视频的质量。因此,有关屏幕内容图像或视频高质量恢复的研究就显得尤为重要。论文基于HEVC框架,重点研究了屏幕内容视频的高质量恢复问题。研究内容主要包括:第一,针对HEVC重建的屏幕内容图像存在畸变和失真的问题,提出一种改进的三维块匹配滤波器(Blcok-matching and 3D Filter,BM3D)作为HEVC屏幕内容的后处理滤波器。通过分析屏幕内容及其噪声的特点,首先对块进行分类搜索匹配,使得匹配组中有数量足够多的匹配块;然后对屏幕内容块进行分割,在更灵活的尺寸上分别进行匹配,使得组中的匹配块具有更强的相关性;最后充分利用图像的相关性进行协同滤波并聚合成去噪的图像。实验结果表明该算法不仅可以提高图像峰值信噪比,对于主观质量也有明显的提升。第二,针对已有的HEVC环路滤波对复合屏幕内容视频的滤波效果不足的问题,提出一种改进的自适应环路滤波器(Adaptive Loop Filter,ALF)作为HEVC复合屏幕内容的环路滤波器,通过对ALF中基于块的自适应分类方法(Block-based Adaption,BA)的改良提高编码质量。实验结果表明,相对于HEVC SCM-7.0,该算法对HEVC复合屏幕内容在All intra,Low Delay,Random Access三个配置下分别能带来平均1.03%、1.65%、1.33%的 BD-rate 的降低。
[Abstract]:Screen content, image or video are widely used in video conferencing, remote desktop control, wireless display and other devices. Unlike the natural content captured by a camera, screen content usually contains a large amount of text, charts and artificial content, and traditional video coding tools cannot compress it very efficiently. To this end, the Joint Collaborative Team on Video coding working Group (JCT-VCC) is committed to developing standard screen content extension tools for efficient video coding High-Efficiency video coding, such as intra-frame block copy mode, palette mode, adaptive color domain transformation, etc. The HEVC screen content extension standard, Screen content coding, was released in February 2016. Although SCC can improve the efficiency of screen content coding, under some conditions, the reconstructed screen content image or video still has a lot of noise and distortion, which reduces the quality of reconstructed image or video. Therefore, the research of screen content image or video quality restoration is particularly important. Based on the HEVC framework, this paper focuses on the high-quality restoration of screen content video. The main contents are as follows: first, aiming at the distortion and distortion of HEVC reconstructed screen image, an improved 3D block matching filter Blcok-matching and 3D filter BM3D is proposed as a post-processing filter for HEVC screen content. By analyzing the characteristics of screen content and its noise, the block is first classified and matched so that there are enough matching blocks in the matching group, and then the screen content block is segmented to match separately in more flexible dimensions. The matching block in the group has stronger correlation. Finally, the correlation of the image is used to filter and aggregate to the denoised image. The experimental results show that the proposed algorithm can not only improve the PSNR of the image, but also improve the subjective quality. Secondly, aiming at the problem that the existing HEVC loop filter is not effective enough for the video content of composite screen, an improved adaptive Loop filter is proposed as the loop filter for the content of HEVC composite screen. The coding quality is improved by improving the Block-based adaptive classification method in ALF. The experimental results show that compared with HEVC SCM-7.0, the algorithm can reduce the BD-rate of HEVC composite screen by an average of 1.33% (1.65%) under the All intraocular low DelayRandom Access configuration.
【学位授予单位】：北方工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN919.81;TP391.41

【参考文献】