Research On Video Coding Algorithm For Visual Perception

With the development of broadband networks and high-definition displays,people have higher expectations for the quality of video pictures,which also brings new requirements and challenges to video coding technology.The rate-distortion optimization technology can guide the encoder to select the optimal encoding parameters and is widely used in video encoding processes.Accurately assessing encoding distortion is the key to rate-distortion optimization technology to reasonably allocate bit resources and achieve optimal compression performance.The distortion estimation method used by the current mainstream ratedistortion optimization techniques only evaluates the degree of damage to the reconstructed signal by only counting the difference in pixel values between the reconstructed signal and the original signal.The human visual system is the target consumer and final receiver of the video.Distortion evaluation methods that do not consider the perception mechanism of the human visual system cannot accurately reflect the subjective perception of the video quality of the human visual system.Establishing a perceptual distortion measurement model that conforms to the rules of visual perception is of great significance to guide the encoder to optimize the allocation of bit resources and improve the compression effect of the encoder.Based on the human visual perception mechanism,this thesis establishes a new perceptual distortion measurement model through objective quality assessment model VMAF and subjective experiments,and applies it to the H.266/VVC encoding standard encoder VTM.The major contributions of this thesis are summarized as follows:An image coding algorithm for visual perception is propesed by considering the influence of spatial complexity on visual perception.Based on HVS visual sensitivity and the results of the VMAF model's evaluation of image quality,this thesis establishes a new perceptual distortion measurement model for region of interest through analyzing the influence of spatial complexity on visual perception.Based on the visual attention mechanism of the human visual system,and combined with the visual importance weights of non-interest areas obtained from subjective experiments,a perceptual distortion measurement model for non-region of interest is established.Applying the established model to the encoder VTM improves the encoder's allocation strategy for bit resources.Experimental results show that compared with the original VTM algorithm,the proposed algorithm has better perceived quality of the reconstructed image under the same bit resources.A video coding algorithm for visual perception is proposed by considering the influence of the spatial and temporal complexity on visual perception.Based on the image perceptual distortion measurent model,the impact of temporal complexity on visual perception is analyzed when the spatial complexity is constant,and a video perceptual distortion measurent model is established and applied to the encoder in VTM,which guide the encoder to optimize the allocation of bit resources.Experimental results show that compared with the original VTM algorithm,the proposed algorithm effectively improves the perceived quality of the reconstructed video under the same bit resources.