Research On Key Technology Of Image Compression Algorithm And Implementation Based On Human Vision System Characters

The rapid development of computer and semiconductor technology leads the great leap forward of the information era.In the information era of big data,image data accounts for a considerable proportion.While pursuing high resolution and high frame rate,it is also confronted with the challenge of mass data transmission and storage.For this reason,many image coding standards have been proposed one after another,among which the HEVC(High Efficiency Video Coding)standards issued by JCT-VC(Joint Collaborative Team on Video Coding),has excellent performance in both video and image coding.Considering that the human eyes are the final receiver of most image information,the traditional coding method does not consider the HVS(Human Visual System)characteristics.There will be a lot of visual redundant information left in image coding,and cannot give full play to the best performance of HEVC coding standard.Therefore,integrating the HVS characteristics in image coding has theoretical research significance and practical application value.The goal of this study is to obtain the best perceptual quality encoded image with limited storage space and transmission bandwidth.The focus of this study is to realize how to combine HEVC encoding standard and HVS characteristics to obtain the best image encoding performance.Based on the HVS characteristic models(including just noticeable difference and region of interest models),a mathematical sense of pixel perception distortion measure is defined to estimate the perceptual distortion of image pixels.The defined pixel perceptual distortion evaluation method was transplanted to HEVC instead of the default objective pixel distortion evaluation method,and the association between HEVC coding standard and HVS characteristics is created.The combination of HEVC coding standard and HVS characteristics can reduce the visual redundancy of the image during the coding process and achieve better image coding performance.In order to realize the goal of joint coding,the two key parameters distortion and bitrate of image coding are taken as the entry points to find the corresponding solutions.On the one hand,if the image perceptual distortion is the same,the bits of image coding should be reduced as much as possible.On the other hand,if the image coding bits are the same,the image perceptual distortion should be reduced as much as possible.From the perspective of HEVC image coding theory,the understanding of the problem can be summarized as: solving the rate distortion optimization problem caused by the introduction of perceptual distortion under different conditions.Based on rate distortion optimization theory,the key to the problem can be further targeted at the determinant of rate distortion optimization,the Lagrange multiplier λ.The research work of this topic is based on the relationship between HVS model and perceptual Lagrange multiplier.The main work and innovations include:On the conditions of specifying the quantization parameters: For the JND models,two methods are proposed to obtain perceptual Lagrange multipliers.One is to establish a look up table of Lagrange multiplier coefficients that is independent of the image content and related to the pixel distortion threshold value by studying and analyzing the correlation between coding parameters with and without considering human visual system characteristics.In image coding,the perceptual Lagrange multiplier is calculated according to the weighted formula related to the image content.The other method is to add pre-processing operation.After the pre-processing operation,the required image distortion and bitrate data related to the image content were got,and then the perceptual Lagrange multiplier value can be calculated according to the rate distortion optimization theory.Based on the rate distortion theory,this paper analyzes how to select the appropriate perceptual Lagrange multiplier when quoting the ROI model,and proves the feasibility of using the default Lagrange multiplier.On the condition of constrained bitrate,according to the traditional rate control method for λ domain based on rate distortion optimization theory,this paper puts forward the R-λ-QP relation model of rate R,Lagrange multiplier λ and quantization parameter QP,which is suitable for image perceptual coding.The proposed perceptual model is an extension of the traditional model,and takes the traditional model as a special case of the perceptual model.In order to ensure the accuracy and the rationality of the coding bits,we also propose a bit allocation method based on the coding tree unit,which achieves the goal of reasonable allocation of the coding bits according to the pixel significant value.In order to test the coding performance for the proposed algorithm,all algorithms are implemented on the HEVC algorithm reference software HM(HEVC test Model).The experimental results show that the proposed solution for different constraints improves the performance of image perceptual coding in different degrees,which proves the effectiveness and correctness of the proposed method.At the same time,considering the practical application of image perceptual coding and taking advantage of the speed of hardware parallel processing,this paper implements the modelsim based functional simulation test under the condition of specified quantization parameters based on the open source HEVC coding core and the JND distortion models,which verifies the feasibility of the proposed method in engineering.