Fast Algorithm Of HEVC Intra Coding And Its Preprocessing Structure

High Efficiency Video Coding(HEVC)was born in 2013,developed by ITU-T and ISO/IEC.It is an upgraded version of the previous generation coding standard H.264/AVC.The compression ratio can be doubled under the same compression quality.However,H.264 is still the most widely used in the market.The development of HEVC is limited by many situations.One of the most important points is that its encoding algorithm is too complicated.Take intra prediction as an example,the prediction mode of HEVC is increased from 9 to 35 compared to H.264.At the same time,the size of the coding unit is expanded,a flexible quadtree partition structure is adopted.HEVC iteratively traverses all depths and prediction modes to select the optimal division,which brings a huge amount of calculation.Therefore,the research on the optimization of intra prediction is very important.Due to the complexity of the coding algorithm,the CPU-based software encoder can not meet the real-time compression of ultra-high-definition video.It is necessary to construct a dedicated hardware encoder.In this paper,a hardware-oriented intra-prediction preprocessing algorithm is proposed.The optimization algorithm mainly includes two aspects,the optimization of the coding unit size decision and the reduction of the prediction mode calculation.In this paper,the prewitt operator is used to extract the texture direction of the highest depth 4×4 size coding block,and the texture complexity of the large size coding block is obtained through statistical analysis,so as to judge whether the size of the coding block can be determined in advance.At the same time,the pixel has stronger similarity along the texture direction,so the prediction mode of the direction has a greater possibility to be selected as the optimal mode.In this paper,the prediction modes are divided into four groups in four directions.Selecting the corresponding mode group according to the significant texture direction enters the RMD phase,effectively reducing the number of mode selections.The above intra prediction optimization algorithm was implemented on HM16.7.A total of 19 test sequences at different resolutions were tested.The coding time of the optimization algorithm can achieve 55.1% reduction compared with the original full scan algorithm,with only 1.35% average bit rate(BD-BR)rise and-0.07 peak signal-to-noise ratio(BD-PSNR)loss.Compared with other recent intra prediction optimization algorithms,the algorithm in this paper has been greatly improved.At the same time,this paper designs a hardware module that preprocesses the original pixels.The module can realize the texture direction calculation of the 4×4 basic block,statistically analyze the texture direction of the large-size CU,and make the advance size decision.This pre-processing module can be easily embedded in other optimization algorithms and meet the throughput requirements of 8K×4K@73 fps real-time preprocessing.