Research On HDR Video Coding Method Based On Visual Perception

Compared to traditional low dynamic range video,High Dynamic Range(HDR)video delivers richer detail and contrast information in highlights and darker areas,giving people a more real-world visual experience.Therefore,HDR technology has broad application prospects in the fields of surveillance security,film and television entertainment.Although the existing High Efficiency Video Coding(HEVC)standard is compatible with coding of High Dynamic Range(HDR)video,considering the particularity of high dynamic range and high precision of HDR video as well as the complex characteristics of the human eye's perception,there are still a lot of perceptual redundancies in the encoded video.Based on the Human Visual System,this Paper mainly about research on the rate-distortion optimization of HEVC and the pre-prossing technology of HDR video coding system and the compatibility of video streams work for different dynamic range users.The final purpose is removing the perceived redundancy while preserving the video quality.Due to the existing rate-distortion optimization technology of HEVC lacks consideration of the perceptual characteristics of the human-visual system to the coding distortion of HDR video,an HDR video coding rate distortion optimization algorithm based on the contrast sensitivity function is proposed.First,according to the definition of Contrast Sensitivity Function,the relationship between Contras Sensitivity and the human eye's sensitivity to errors is obtained.Then,for the luminance component of 10-bit HDR video,Coding Tree Unit(CTU)is taken as the basic unit to get the adaptation luminance value and the spatial frequency which affect the visual contrast sensitivity.Finally,the contrast sensitivity value is further corrected as the visual perception factor.The Lagrange factor of CTU is scaled to achieve a more efficiency assignment of the code rate consistent with visual perception and remove the perceptual redundancies.Experimental results show that the code rate is 5.6% and 8.3% lower than that of the original coding system with the same Q-factor of HDR-VDP-2 and PSNR_DE scores,which shows that the proposed method can effectively reduce the coding rate while maintaining the perceived quality of HDR video.Considering the distortion caused by the character of working for a single dynamic range of the existing pre-prossing technology of HDR video coding system,this paper proposed an adaptive pre-prossing quantizer technology according to the content and dynamic range of HDR video.Firstly,we analyze the contrast distortion of HDR video caused by the Perceptual Quantizer(PQ)in pre-prossing.Then,according to the distribution of histogram of each channel luminance of the HDR video and the contrast perception characteristic of human visual system,the distribution of the 10-bit integer quantized signal is optimized to protect the contrast information of the HDR video as much as possible;Finally,the original HDR video is optimized by mapping according to the optimization strategy combined with the inverse PQ,to achieve an adaptive pre-processing quantization and reduce the distortion of HDR video.Experimental results show that the code rate is 8.7% and 15.3% lower than that of the original coding platform with the same Q-factor of HDR-VDP-2 and PSNR_DE scores,which shows that the proposed method can effectively improve the performance of HDR video coding system.Aiming at the problem of the compatibility of video signal streams work for Standard Dynamic Range and HDR users,this paper proposes two dynamic range scalable HDR video coding schemes based on the existing Tone Mapping Operate(TMO).In scheme 1,we analyze the HDR distortion information contained in the residual layer combination with the visual perceptual characteristics,and the residual layer is processed by the method of guided filtering to reduce the code rate generated by the coding system.In the second scheme,we analyze the TM method firstly,and then extract the TMO for different HDR video frames.The TMO is further modified and transmitted as auxiliary information to the decoding part and reconstructs the HDR video.Compared with the encoding method of transmitting residual layer,TMO can save a certain code rate as auxiliary information and reduce the distortion caused by the compression algorithm due to the characteristics of single layer and have low frequency information,while verifying the effectiveness of the dynamic range scalable coding method.