Encoding Optimazation Of A Scalable Mezzanine Codec SUVC

Video coding,which has been developed for decades,is a technology utilizing redundant information of video to implement data compression.Today,with the rapid development of 5G,ultrahigh-definition（UHD）videos such as 4K and 8K are widely popularized,and more and higher quality UHD content is arriving in the production environment of industry media,resulting in much more bandwidth and storage cost.Therefore,adding a low-complexity,low-latency,high-quality mezzanine codec to the video production workflow can effectively decrease bandwidth and storage consumption,and even making transmiting or producing higher-quality video content on old infrastructure possible.In this thesis,we mainly propose a scalable mezzanine codec SUVC（Scalable Ultradefinition Video Coding）,and the transform,bit rate allocation,and entropy coding module of the enhancement layer of SUVC are analyzed,which improve the coding efficiency of SUVC.The work of this thesis can be listed as follows:（1）We propose a scalable mezzanine 2D-DWT based codec,which is SUVC.The SUVC possesses the characteristics of low latency,low complexity,high quality,and scalability.Besides the reconstruction quality of SUVC is better than the latest mezzanine codec,JPEG XS,and widely used codec in industry media,XAVC.（2）For the entropy coding module in the SUVC enhancement layer coding,this theis analyzes the probability distribution of transformed and quantified coefficients of the enhancement layer of SUVC.For sources entropy is less than 1,quadtree significance coding is proposed to improve the coding efficiency.For sources entropy is greater than 1,an improved adaptive Golomb coding is proposed on the basis of Golomb coding in terms of the deriving and analyzing of probability distribution of quantization coefficients.Compared with 0th exponential Golomb encoding,the improved adaptive Golomb encoding achieves an average BD-rate gain of 5.57%,5.45%,and 5.47% of YUV respectively for 4K video,and achieves an average BD-rate gain of 4.86%,4.89%,4.93% of YUV respectively for8 K video.（3）For the transform module of the SUVC enhancement layer,this thesis proposes to use the 2 × 2 Hadamard transform for the enhancement layer coefficients,and then studies the code efficiency of Hadamard transforms of different sizes.Therefore,this thesis proposes a content-adaptive Hadamard transform method,which has higher compression efficiency for video with few color and achieves up to 44.06% BD-rate saving on V-PSNR compared with 2 × 2 Hadamard transform applied in SUVC only.（4）For the bit rate allocation module in SUVC enhancement layer,a heuristic variable bit rate allocation scheme is proposed according to the probability distribution of the enhancement layer coefficients.Compared with average bit rate allocation method,it achieves an average BD-rate saving of 4.53%,6.75%,and 5.73% of YUV respectively for 4K video,meanwhile it can achieve an average BD-rate saving of 11.02%,11.27%,10.77% of YUV respectively for 8K video.