Research On Characteristics-based Intra Coding Optimization Algorithm For HEVC-SCC

With the emerging requirements such as cloud gaming,remote desktop interfacing,and video conferencing,screen content video(SCV)has gained more and more attention as a special video type.In terms of video encoding for screen content,video coding expert group(VCEG)and motion picture expert group(MPEG)jointly launched the development of the HEVC-based screen content encoding extension standard in January 2014,and in 2016 developed HEVC-SCC,an extension of the high efficiency video coding standard.Unlike traditional camera-captured video(CCV),SCV is usually generated by computer,including text,graphics,user interfaces,or a mixture of camera captured content and computer-generated content.Since the previous algorithms are mainly designed for CCV,they do not consider the characteristics of SCV,such as a large number of flat areas and sharp edges,which reflect poor improvement for a mixture of nature content and screen content if directly applying.In response to the above problems,the characteristics of the screen video are analyzed in depth,and some low-complexity HEVC-SCC fast coding algorithms are proposed to reduce the computational complexity for screen content video encoding while maintaining the video quality.The main tasks are as follows:Firstly,a fast intra prediction algorithm for HEVC-SCC based on the content property and texture mode is proposed.Since DCT coefficients energy distribution of screen content video is differ from the natural content ones,early decision of video content is helpful to skip some prediction modes.With gradient information of prediction unit(PU),coding tree units(CTUs)can be classified into natural content CTUs(NCCTU),smooth screen content CTUs(SSCTU)and complex screen content CTUs(CSCTU).For NCCTU,traditional 35 intra modes are selected as candidates,and the Intra Block Copy(IBC)and Palette(PLT)modes are skipped;for SSCTU,DC,PLANAR,horizontal and vertical modes are chosen as the best modes,while the judgment for SCC are skipped;for CSCTU,the IBC and PLT modes are used for prediction and skip traditional intra modes.Under all-intra(AI)configuration,experimental results show that the proposed algorithm can achieve encoding time saving by 38.55% compared with SCM-8.3,and the bit rate only increases 1.82%.Considered that characteristics of SCV are reflected in large flat areas and sharp edges,a fast algorithm based on gray level co-occurrence matrix and Gabor feature model for HEVC-SCC,denoted as GGM,is proposed in this paper.The algorithm includes fast mode decision and CU size decision.By studying the correlation of nonzero number in GLCM with different partitioning depths,the coding unit(CU)size of intra coding can be prejudged,which selectively skips the intra prediction process of CU in other depths.With Gabor filter,the edge information reflecting the features of screen content images are extracted.According to Gabor feature,CUs are classified into different types,with which,the calculation and judgment of unnecessary prediction modes are skipped.Under AI configuration,experimental results show that the algorithm can achieve encoding time saving by 42.13% compared with SCM-8.3,and with only 1.85% bit-rate increasement.