| AVS2(Audio Video Standard II)video coding standard is the second generation of independent intellectual property coding standard which is independently developed by our country.It is also a new coding standard formulated by the AVS Working Group on the basis of AVS and AVS+.AVS2 was consummated in June 8,2015 and comparable to the international standard HEVC(High Efficiency Video Coding)at the same time,indicating that our country’s technical development in the field of video coding has reached the international advanced level.While the new coding standard brings advanced coding tools and technology,video coding compression ratio to get doubled enhance,at the same time it also brings dramatic increase in coding complexity,it is difficult to meet the needs of real-time encoding,so it is very necessary to do some research deeply in AVS2 video encoding optimization algorithm.Firstly,this paper introduces the framework and structure used in video coding of AVS2,and then introducing the key technology of AVS2 video coding in detail.Studies have shown that in the all-intra mode,intra prediction of AVS2 takes the first place and has a rather high coding complexity.This paper proposed a fast intra prediction mode algorithm based on CU texture classification after intensive research in AVS2 intra prediction process,according to the texture complexity,the types of CU are classified.At the same time,the complete set of33 intra prediction models is divided into 5 types of prediction patterns based on the classification of CU,and the mapping relationship between the prediction pattern set and the CU type is established,reduced the number of the prediction modes required to traverse,thus,the coding efficiency has been improved,and the purpose of algorithm optimization is achieved.Experiments show that the new algorithm reduced the encoding time by 48%~52%at the cost of 2.84% bit rate increase and 0.13 dB Peak Signal to Noise Ratio(PSNR)loss on average compared to all-intra modes at the AVS2 reference software(RD 14.0).Then,according to the same research method,we change the scope of action and proposed a fast intra prediction mode algorithm based on PU texture classification after intensive research in AVS2 intra prediction process.Experiments show that the new algorithm reduced the encoding time by 44%~48% at the cost of 4.81% bit rate increase and0.11 dB Peak Signal to Noise Ratio(PSNR)loss on average compared to all-intra modes at the AVS2 reference software(RD 14.0).Then,the existing intra prediction optimization algorithm of the project group is studiedand analyzed.Document [45] is different from the proposed algorithm in the intra prediction process.it is possible to combine the Document [45] with the proposed algorithm.It’s verified by experimental and simulation data.The joint algorithm can enhance the coding efficiency by nearly 70%,which shows the superiority of the joint algorithm. |
