Research On Fast Algorithm For HEVC Screen Content Coding

With the emergence of video conferencing,wireless presentation,remote education and other application scenarios,computer-generated screen content is playing a more and more important role in people's daily lives.Compared with the natural content captured by cameras,screen content has totally different characters like containing graphic texts,shape edges and limited colors.As the latest video coding standard,High Efficiency Video Coding(HEVC)can perform well when processing natural content.However,large amount of data will be generated when HEVC handles screen content.To solve this problem,the HEVC expansion Screen Content Coding(SCC)is proposed.In SCC,new coding tools like Intra Block Copy(IntraBC)and Palette mode are integrated in the platform,which greatly improves the coding efficiency toward screen content.On the other hand,new tools bring more computation burden to the encoders,and much more time is consumed during the coding process,which becomes an obstacle for SCC to be applied in low-latency and real time situations.To reduce the computational complexity of SCC,a fast SCC inter mode selection algorithm is proposed in this paper.First,the change of luminance values between current pixel and its co-located pixel is utilized to detect static regions and Skip mode is applied.Next,given that a large number of horizontal and vertical edges occur in screen content,corresponding prediction unit(PU)partition modes are selected according to coding unit(CU)activity.Finally,depth range of current CU is predicted by utilizing the CU depths of adjacent CUs in both temporal and spatial domains,and unnecessary depths are skipped.Experimental results show that,compared with the anchor SCC algorithm,the proposed algorithm can save 43.56% and 49.09% encoding time on average with 3.06% and 3.43% BD-rate loss under random access and low delay configurations respectively,while the loss in video quality is negligible.In the fast SCC intra mode selection algorithm,luma values of all pixels in both natural content and screen content are counted,and video content is classified according to their different statistical features.For natural content,IntraBC and Palette modes are skipped,and best intra mode for current CU is predicted by utilizing the best mode information of adjacent CUs.For screen content,given that a large amount of background areas are simple textured,the whole process of Rough Mode Decision can de skipped,and corresponding prediction modes are selected for different kind of areas.Thus,the number of modes for full Rate Distortion Optimization process is reduced,and fast SCC algorithm is achieved.Experimental results show that compared with the anchor SCC algorithm,the proposed algorithm can reduce 37.57% encoding time on average with 2.08% increase in BD-rate and 0.19 dB decrease in PSNR.