Optimization And Implementation Of Quantization Algorithm For Video Coding

The H.264/H.265 video coding standard uses the hybrid coding framework,including prediction,transform,quantization,entropy coding,loop filtering modules.Quantization module plays an important role in the hybrid framework,which directly determines the distortion and coding rate.The module of the quantization mostly adopts the hard decision quantization(HDQ)algorithm based on fixed dead-zone.In order to meet the higher rate distortion performance requirements.The module of the quantization adopts the rate distortion optimization quantization(RDOQ)as candidate quantization algorithm,RDOQ is a kind of soft decision quantization(SDQ)algorithm based on the rate distortion theory.SDQ algorithm achieves superior coding performance,however suffers from deadly sequential processing dependency.Comparatively,HDQ supports high-level parallel processing and is dependency-immune,however suffering from non-negligible coding performance degradation.Based on the present situation of quantization algorithm,this paper proposes an adaptive deadzone offset model to improve the traditional HDQ algorithm.The algorithm takes into account both the advantages of soft and hard decision quantization,and reduces the rate distortion performance loss as far as possible on the basis of the parallel processing.It is found that the distribution characteristics of the transform coefficients and the quantization parameters have a great influence on the quantization,and the number of non-zero coefficients in the neighboring blocks will have a fine tuning effect on the quantization of the current coefficients.coefficient-wise offset model is built as the function of three parameters.This model is elaborately design by maximizing the right judgment probability and minimizing the wrong judgment probability to improve the coding performance as much as possible.Based on the model,we propose an adaptive deadzone quantization algorithm for H.264 coding standard.In order to extend the adaptive dead-zone quantization to the latest video coding standard and considering the human visual perception,this paper proposes a perceptual adaptive hard decision quantization algorithm in a late stage study.In this paper,under the H.264 standard,the fixed dead-zone quantization algorithm,the adaptive dead-zone quantization algorithm and the rate distortion performance of the rate distortion optimization quantization algorithm are verified by simulation experiments.Compared with the fixed dead-zone quantization algorithm,the HDQ algorithm based on the adaptive offset model of this paper obtains significant performance improvement,PSNR has 0.08836dB upgrade,equivalent to 3.097%bit rate savings.Compared with SDQ algorithm,the performance of PSNR is only 0.03921dB,which is equivalent to 1.51%bit rate increase.The algorithm complexity is consistent with the traditional hard decision quantization algorithm.The rate distortion optimization algorithm proposed in this paper is verified by the H.265/HEVC standard,and the rate distortion performance and complexity of the rate distortion optimized quantization algorithm are verified.Compared with the traditional rate distortion optimized quantization algorithm,the algorithm proposed in this paper,BD-RATE has an average of 2.6094%saving rate,BD-PSNR average performance improvement of 0.0521dB.The coding time is saved 4.3305%.The perceptual adaptive hard decision quantization algorithm which is considered the subjective perception characteristics of human eyes,in the subjective evaluation test,the algorithm is close to the perception RDOQ algorithm in the subjective perception rate distortion performance.