Research On Low-complexity Ultra-high Definition Video Coding Method

With the massive development of multimedia technology and video-capable devices,people put forward higher requirements for the visual experience of watching video.Meanwhile,the video presents diversified development in content.Ultra High Definition(UHD)video services have been widely utilized in various fields due to their ability to provide users with a high-quality immersive experience and more realistic visual enjoyment.However,the higher video resolution(up to 4K and 8K)has increased the amount of video data enormously,thereby introducing new challenges to video transmission.In addition,the high encoding complexity severely limits the practical application of UHD video,especially for some real-time applications and power constrained devices.Therefore,it is necessary to address the issue of huge computational complexity in UHD video coding.In this paper,three fast coding algorithms have been proposed to reduce the encoding complexity of UHD videos with different video content.(1)A joint video enhancement and fast intra encoding algorithm for depth video coding.Firstly,a depth video enhancement method is introduced to remove the inaccurate textures in depth video,which provides the optimization space for accelerating the encoding processing.The enhanced depth video is mainly characterized by sharp object edges and large areas of nearly constant regions.Then,Coding Units(CUs)are classified into two classes: simple CU and complex CU,according to texture complexity.The CU size decision of simple CUs is early terminated.Moreover,Prediction Units(PU)are classified into two classes according to the edge intensity of PU.The Depth Modeling Mode(DMM)decision in PUs with low edge intensity is skipped.The experimental results show that the proposed algorithm can significantly reduce the encoding time with lower bitrate.Specifically,the encoding time reduction obtained by the proposed algorithm compared with that of the original encoder is 62.91% on average.In terms of encoding efficiency,the proposed algorithm can reduce bitrate by 4.63% under the condition of same synthesized virtual view quality.(2)A fast intra coding algorithm based on neural network and gray co-occurrence matrix for Screen Content Coding(SCC).Firstly,a neural network-based CU classification model is designed to classify CUs into SCCUs and CCCUs.Then,in order to speed up the process of SCC intra mode decision,an efficient PU mode assignment method is proposed to eliminate the computational redundancy derived from unnecessary modes checking.In this way,the candidate intra modes are adaptively assigned according to the type of current CU.Furtherly,we utilize GLCM to evaluate the texture complexity of CUs,and terminate the process of CU size decision in advance.In addition,a weighted factor online updating method is introduced to cope with different characteristics of test sequences.In order to achieve a good trade-off between complexity reduction and RD performance,extensive experiments are conducted to select the optimal threshold.The experimental results show that in comparison with the original SCC reference software,the proposed overall algorithm can reduce 49.33% intra coding time with 1.36% BDBR increase.(3)A multiple classifier-based fast CU decision algorithm for intra coding in Future Video Coding(FVC).The proposed multiple classifier-based Quad-tree Plus Binary Tree(QTBT)algorithm contains three stages including horizon binary-tree decision model(HBTDM),vertical binary-tree decision model(VBTDM),and quad-tree decision model(QTDM).Firstly,three multiple classifier-based QTBT partitioning mode decision models(HBTDM,VBTDM and QTDM)are designed for three different partitioning modes of QTBT structure to speed up the process of QTBT partitioning.Then,in order to balance the RD performance and computational complexity,extensive experiments are conducted to select the optimal training parameter.Finally,the experimental results show that in comparison with the original FVC reference software,the proposed overall algorithm can reduce 65.77% intra coding time with negligible degradation of RD performance.