Research On Perceptual Optimization For High Efficiency Video Coding

As the high-definition video technology like 4K and HDR grows rapidly,consumers' demand for higher-resolution,more immersive video content continues to increase,forcing the video industry to continuously develop more and more efficient video coding technologies.How to make full use of the limited bandwidth resources to encode video streams with higher quality has always been the focus of video coding optimization.Rate control and rate-distortion optimization,as key technologies for video coding and transmission,play a vital role in practical video applications.Specifically,when encoding video sequences,the rate control and the rate distortion optimization modules can make the encoded video stream satisfy a given bandwidth limitation by selecting appropriate encoding parameters or encoding modes,and at the same time make the resulting distortion as small as possible.In order to make the allocation of bitrate in rate control and mode selection more reasonable and more consistent with the human visual characteristics,that is,make the subjective visual quality of the encoded video much better,taking into account some important human visual characteristics and image content characteristics provides a new idea for optimizing the bit allocation and selecting the best coding mode in video coding.This is also a research hotspot in video coding optimization.Aiming at the lack of human visual characteristics or not fully considered in the existing rate control models and schemes,this paper proposes a rate control scheme based on visual masking effects.The scheme not only considers the original way of assigning the bitrate according to the coding complexity,but also takes into account the vital human visual characteristics,and then the content perception factor of the local region is derived according to the texture complexity and motion intensity of the local image region to be encoded.The proposed content perception factor can effectively characterize the visual importance of the current image region,and thus can be used to provide guidance for assigning bitrates to each coding unit in the frame,like allocating more bitrate resources to more visually important regions.The proposed scheme can make the quality of the encoded video more in line with the human visual characteristics,that is,the subjective visual quality becomes much better at the same bit rate.And finally,the average BD-Rate gain based on MOSp can achieve almost 4.49%.The implementation details of the optimization scheme are described in detail in the subsequent sections of this paper,and experimental results are provided to evaluate the performance of this scheme.Aiming at the lack of human visual characteristics for the existing ratedistortion optimization schemes,this paper proposes a new rate-distortion optimization scheme based on visual characteristics.Firstly,the proposed scheme improves and obtains a pixel-domain JND model based on the visual characteristics like luminance masking and contrast masking.And then for the rate-distortion optimization process,a novel Lagrangian multiplier adjustment coefficient depending on the visual importance of the image region will be deduced according to a proposed perceptual distortion model.The refined Lagrangian multiplier used in the adaptive dynamic ratedistortion optimization process can effectively reduce the bitrate resource consumed by the encoding process,meanwhile maintaining the subjective quality substantially unchanged.Finally,the average bitrate reduction can achieve almost 13.1%.The implementation details of the optimization scheme are described in detail in the subsequent sections of this paper,and experimental results are provided to evaluate the performance of this scheme.