Research On Rate Distortion Optimization In AV1 And X264

In 2018,Alliance for Open Media launched the AV1 new-generation video coding standard to promote the development and application of video coding technology.However,in practical applications,the H.264 standard still occupies a large market,and X264 is the excellent H.264 standard encoder.In video coding and decoding technology,ratedistortion optimization technology has played a key role in improving the efficiency of video coding,but the rate-distortion optimization technology of existing encoders is more based on the optimization of the coding mode of the current coding block.Considering the influence on the coding mode of subsequent coding blocks,the time-domain dependence characteristic is not fully considered when solving the rate-distortion optimization problem.This paper mainly studies the time-domain dependent rate-distortion optimization problem of AV1 standard and X264 encoder,and mainly optimizes and improves related technologies:(1)The time domain dependency rate distortion optimization algorithm is studied for the default coding structure of AV1 video encoding standard,that is,random access hierarchical coding structure.By analyzing the time domain hierarchy dependency of AV1,establishing the time domain propagation chain,and under the strict deduction of mathematical model,estimating the degree of distortion propagation,solving the propagation factor of coding blocks,and realizing the optimization of global rate distortion of time domain dependence.In addition,the relationship between the quantitative parameters of AV1 and the Lagrangian multiplier is analyzed and the relationship formula is revised.The algorithm is applied to the AV1 test platform libaom,and the standard video sequence test is used to achieve 1.59% coding gain and 7.9% code time saving.(2)In view of the rate distortion optimization problem of X264 encoder,a real-time time domain-dependent rate distortion optimization method is proposed under the lowlatency B-frame and low-latency P-frame coding structure.This method pre-reads 1frame of encoded content,establishes the time domain propagation chain only uses adjacent frames,and optimizes the I frame according to the characteristics of the low code rate and strong dependence on distortion,which is more in line with the practical application of the real-time coding scene.This method is applied to the X264 encoder and tested for pre-reading all video frames and 1 frame.Experimental results show that the method of pre-reading 1 frame does not increase the complexity of encoding time,and under the low-latency B-frame coding structure,the method of pre-reading all frames and1 frame has 12.26% and 10.25% coding gain respectively.The method of pre-reading 1frame is guaranteed while ensuring the application of the actual scene,and the coding performance is guaranteed compared to the pre-read all frames.