Visual Perception Based Optimization Method For High Efficiency Video Coding

With the development of digital media technology,the amount of video data is largely increased.How to efficiently reduce this data and maintain the quality has become a challenging task.To address this challenge,ITU-T and ISO/IEC have been working together to develop effective video coding standards.At present,the latest video coding standard,i.e.,High Efficiency Video Coding(HEVC),has doubled the coding efficiency of its predecessor,i.e.,H.264 / Advanced Video Coding(AVC).However,the human eye is the ultimate receiver of video data.Combining the perceptual characteristics of human visual system with video coding is able to remove the visual perception redundancy and bring more coding gains.Based on this,in this thesis,the characteristics of human visual perception have been introduced into the video coding framework,and a visual perception based optimization method for HEVC has been presented.The main contributions of this thesis are listed in the following aspects:1.In HEVC,the Coding Tree Unit(CTU)level bit allocation algorithm for Intra frame uses the Mean Squared Error(MSE)or Mean Absolute Difference(MAD)of co-located CTU as a weight to guide bit allocation.This approach does not fully consider the visual perception characteristics of the human eye.Therefore,a bit allocation algorithm based on spatial domain visual perception distortion is proposed.First,replace the traditional distortion metric with the perceptual distortion metric that incorporates the spatial domain masking effect,and establish a new model of the relationship between the spatial domain perceptual distortion and the rate.Next,verify the accuracy of the model through data statistics and fitting.In addition,two parameter estimation methods are proposed to solve the parameter model.Finally,the perceptual weighting factor guide bit allocation of each CTU is solved by deriving the rate-distortion optimization theory and the Lagrangian multiplier method.In order to verify the performance of the proposed algorithm,a spatial domain perceptual evaluation model is used to evaluate the rate distortion performance of the proposed algorithm.Besides,seven other video qualityevaluation models were also used to evaluate the performance of the proposed algorithm.The experimental results show that under the same perceptual quality,the proposed algorithm reduces bit rate by 6.27% on average,which greatly improves the encoding performance.2.For the bit allocation of I-frame images,considering the spatial visual perception characteristics can improve the coding efficiency,but for the encoding of non-I-frame images,only considering the spatial visual perception characteristics of the video is not enough to remove the visual perception redundancy,because the video consists of many frames.When encoding non-I-frame images,the temporal domain perception characteristics of the video also need to be considered.The perception model that incorporates visual features of the spatial and temporal domain can better evaluate the perceptual distortion of non-I-frame images.Therefore,it is proposed to use the perceptual distortion model based on the spatiotemporal domain to guide the rate control technology in video coding.First,the I-frame image in the video is removed,and according to the statistical process to building the relationship model between rate and spatiotemporal perceptual distortion of the non-I-frame image.Then,according to the relationship between spatiotemporal perceptual distortion of the non-I frame and the bit rate,the perceptual weight of each CTU is obtained.After that,CTU-level bit allocation is guided according to the perceptual weight.Finally,the R-? and QP-? models are used to obtain the encoded quantization parameter values,which then guide the CTU-level bit rate control algorithm.Compared with the algorithm of JCTVC-K0103,the proposed algorithm greatly improves the efficiency of video coding.3.In order to meet the image quality required for clinical treatment and to improve the compression ratio as much as possible,a Region of Interest(ROI)based video coding method for pathological slice image is proposed.Firstly,the perceptual characteristics of pathological slice images are used to determine image thresholds,and pathological slice images are divided into ROI and non-ROI.Then determine the perceptual distortion values of the ROI and non-ROI in the pathological slice image.The ROI of pathological slice images is compressed by the human eye without perceiving distortion,and themaximum distortion compression is used for non-ROI without affecting the visual effect.Experiments show that the compression method for compressing pathological slice images by region is a 92% bitrate saving compared to compressing pathological slice images using the same method.In the encoding process,in order to control the perceived quality of different regions,it is necessary to traverse all quantization parameters and calculate the corresponding distortion value,and the calculation complexity is relatively high.By establishing a model of the relationship between distortion and quantization parameters,the selection of quantization parameters can be quickly implemented,which greatly reduces the complexity of encoding.The experimental results that the time-saving rate of the ROI-based pathological slice image fast encoding method is 97% compared with the full traversal encoding method,which not only maintains the compression efficiency of the encoding,but also greatly reduces the encoding complexity.The visual perception-based method for high efficiency video coding optimization fully considers human visual perception characteristics such as masking effects,contrast sensitivity,and visual attention mechanisms.The perceptual model is added to the video coding framework,which eliminates the perceptual distortion of the human eye,optimizes the allocation of bit resources,and effectively improves the performance of video coding.The application of perceptual coding to pathological slice images greatly improves the compression efficiency of pathological slice image.The research of this topic has a certain theoretical basis for the development of perceptual coding,and has a positive guide significance for video conference,image and video compression,media communication,and smart medicine.