| As a new generation of video compression coding standard,H.266/VVC introduces a number of new coding technologies compared with the previous generation of standard H.265/HEVC,which greatly improves its coding performance.However,ithas a high coding complexity.The introduction and expansion of coding technologies lead to the limited optimization effect of some previous coding complexity optimization algorithms on H.266/VVC.Therefore,it is urgent to study H.266/VVC and propose the corresponding complexity optimization algorithms.Inter-frame prediction is the most complex part in the H.266/VVC coding process.Considering of these,in this paper,the inter-frame prediction of the coding standard H.266/VVC is studied,and several inter-frame coding prediction complexity optimization algorithms are proposed.Based on the traditional statistical analysis method,this paper proposes a complexity optimization scheme for the key modules of H.266/VVC inter-frame prediction,i.e.Merge technology,AMVP technology and partition mode decision.Firstly,an affine Merge early termination algorithm based on the motion characteristics of content is proposed for Merge technology.By analyzing the relationship between conventional Merge technology and affine Merge technology,affine Merge technology is skipped when conventional Merge technology can provide good prediction results.Then,for TZSearch motion estimation search,which is the core of AMVP technology,a TZSearch motion estimation search optimization algorithm based on comparison of search path costs is proposed.In the process of motion search,the search path is optimized according to the cost variation trend of each search point.Finally,a CU partition early termination algorithm based on the ratedistortion cost of the encoded sub-block is proposed for the partition mode decision.The algorithm calculates the ratio of the sum of the rate-distortion cost of the encoded subblock to the rate-distortion cost of the parent block in the decision process.When the ratio is greater than the given threshold,the encoding process of subsequent sub-blocks is skipped.In this paper,the proposed algorithm is applied to the VVen C 1.0 encoder.The experimental results show that using the above three optimization algorithms simultaneously can reduce the coding complexity by an average of 6.89% and only lose0.23% of the performance on the basis of the slow level of the VVen C 1.0 encoder.On the basis of optimizing key modules by traditional methods,this paper further designs a multi-level inter-frame prediction optimization algorithm for H.266/VVC based on random forest.In the algorithm,the prediction mode decision process of H.266/VVC is devided into three stages: after the rough selection of Merge technology,after the fine selection of Merge technology and after the default accuracy search of AMVP technology.In each stage,some obtained coding information is used to judge whether the subsequent decision process can be skipped to reduce coding complexity.The judgment content of the algorithm in three stages is as follows: The first is to judge whether other prediction techniques can be terminated in advance according to the rough selection of Merge technology.Then,it is judged whether the subsequent division mode decision can be terminated in advance through the prediction result after the fine selection of the Merge technology.Finally,after the default precision search of the AMVP technology,it is judged whether the multi-precision search and the rate-distortion cost calculation of the optimal mode can be terminated in advance.The multi-level inter-frame prediction optimization algorithm based on random forest is applied to the VVen C 1.0 encoder.The experimental results show that the algorithm can reduce the coding complexity by an average of 22.10%on the basis of the medium gear of the VVen C 1.0 encoder,while the performance loss is 1.61%. |
PDF Full Text Request