Research On Perceptual-based Screen Content Video Coding Optimization Algorithm

With the emergence of new applications such as screen sharing,video conferencing,remote video education and cloud games,higher requirements are put forward for video coding strategies based on specific features of screen content.Unlike the natural video captured by the camera,the screen content video usually has the text-chart area generated by the computer and nature area captured by the camera.In order to provide real-time and low latency transmission for screen content video,ITU-T video coding expert group(VCEG)and MPEG motion picture expert group(MPEG)officially launched the HEVC-SCC standardization process in 2014.Considering that the key rate distortion optimization technology and rate control technology in the original HEVC-SCC coding framework ignore the perceptual characteristics of the human visual system,resulting in low efficiency of perceptual coding.Besides,the content characteristics of screen content video are ignored,such as sharp edges,content invariance and mutation between adjacent frames.To solve the problems,the relationship between human visual perception characteristics and HEVC-SCC rate distortion optimization model is analyzed in depth,which is used to guide the rate control,so as to achieve the purpose of adaptively adjusting the bitrate according to improve the perceptual coding efficiency of screen content video.The main work includes the following three aspects:Firstly,a rate distortion optimization algorithm of HEVC-SCC based on perceptual characteristics is proposed.Considering that the Lagrange operator ? of all blocks in the HEVC-SCC rate distortion optimization model is the same,the perceptual characteristics of video content are ignored.Therefore,the perceptual property can be added to adjust ? adaptively to improve the quality of the reconstructed video and reduce the bitrate.Among them,the extraction of perceptual features mainly includes extracting the spatial and temporal features of current CTU.In the spatial domain,the gradient amplitude information of the screen content is calculated as the spatial domain feature descriptor;in the time domain,the gradient similarity value GS is obtained by combining the gradient direction and gradient amplitude.The new perceptual features are obtained by the weighted fusion of spatial and temporal features.Finally,the obtained perceptual features are added to the HEVC-SCC rate distortion optimization model,the Lagrange operator is adjusted adaptively,and the bit allocation is reasonable,which effectively improves the perceptual quality and rate distortion optimization performance of the video.Based on the first work,we use the spatiotemporal perceptual features extracted to control the bitrate adaptively at the CTU level.Then,the quantization parameters and parameter updating are determined according to the R-? and ?-QP model.Experimental results show that in random access mode,compared with the original HEVC-SCC algorithm,the proposed algorithm can slightly reduce the bitrate mismatch,save the bitrate overhead and improve the video perceptual quality.In view of there are a large number of text and graphics areas,sharp edges and sudden changes of content scene in screen content videos,we proposes an adaptive perceptual rate control algorithm of screen content video.The main idea is that we use the latest screen image quality evaluation method GFM(Gabor feature-based model)as the spatial perceptual feature model.In addition,the temporal perceptual feature model is established to determine the type of the current block.Then spatiotemporal perceptual features are weighted and fused to get final perceptual features,which are used to reasonably allocate bits and update parameters of the screen content video at the CTU level.It can be seen from the experimental results that the proposed algorithm can effectively reduce the bitrate mismatch,improve the quality of video and have better rate distortion performance under two different coding structures.