Research On Quality Enhancement Algorithm Of Compressed Screen Content Video

In recent years,with the rapid development of computer technology,the spread of videos has reached an unprecedented scale.In particular,with the rise of online education,video conferencing,and live gaming,screen content video is getting more attention than ever.At present,efficient lossy video compression technology is adopted to alleviate the contradiction of network congestion caused by massive data and ensure the continuity of playback at the cost of partial video quality loss.It involves the research of video compression technology and how to repair the quality loss caused by the former.Screen content is the combination of text and graphics produced by the electronic screen.At the same time,screen content video can be divided into three types according to the difference of content,which are: mixed content,TGM(text and graphics with motion)and animation.Compared with the video captured by the traditional lens,the screen content video has the following characteristics: no noise,color film,rich and changeable texture.Therefore,since the international Coding standard H.265/HEVC,all compression standards have introduced Screen Content Coding(SCC).Relevant coding modules design compression algorithms according to the characteristics of screen content.Typical algorithms are as follows: Intra Block Copy(IBC),Palette Mode(PLT),Transform Skip Mode(TSM),etc.These advanced compression algorithms will bring great loss to video quality,and there will be a large number of compression artifacts in visual effects.Therefore,thesis mainly studies how to reduce the artifacts of screen content compression video and improve its visual quality.The research content of thesis is as follows:1.At present,mainstream video quality enhancement algorithms do not consider the content features of screen content video,so these algorithms have limited performance in the task of quality enhancement on screen content video.Based on the content analysis results of screen content videos,thesis proposes a video quality enhancement model based on multi-scale difference.It can recover the high frequency information lost by video coding and improve the quality of video viewing.Specifically,the algorithm focuses on the missing areas of high-frequency information and restores them by reference frame information through the difference analysis of feature maps of different scales.2.Thesis designs the quality enhancement algorithm according to the content characteristics of the screen content video.We propose a content adaptive video quality enhancement model based on two branches.Different from existing quality enhancement algorithms,the core point of this model is to carry out adaptive information processing according to content characteristics to select effective information and use it to achieve better enhancement effect,so as to remove artifacts brought by coding as much as possible and improve visual effects.Mainstream video quality enhancement modules mostly perform poorly in screen content compression and video quality enhancement.First of all,there are essential differences between traditional video and screen content video in content,and the content of different categories of screen content video is also very different,so it is more necessary to adopt an adaptive way to deal with such a variety of changes.