Research On Video Coding Algorithm Based On Perceptual Quality

The video coding algorithms cause different degrees of visual quality impairment while reducing storage space and saving transmission bandwidth.The ultimate goal of video coding is to maximize the video stream compression rate under the premise of ensuring the visual quality of videos.Accurately assessing video quality thus become an important prerequisite for optimizing encoding parameter settings and improving coding efficiency.Due to the lack of understanding of the human visual system,early video quality assessment metrics use signal fidelity measures including PSNR to represent video quality.These measures only quantify the pixel difference between the original signal and the distorted signal,and do not consider the impact of the video features and the perceptual characteristics of the visual system.This usually results in a low consistency with the subjective quality perception result of the human visual system.Therefore,establishing a video distortion measure that conforms to the perception law of the human visual system and applying it to video coding has important significance for improving coding efficiency.In this paper,considering the influence of spatial-temporal video features on the video quality perception,an image coding algorithm based on perceptual distortion measurement is proposed on the basis of the latest H.265/HEVC coding standard.Furthermore,considering the impact of temporal video features on video quality perception,a video coding algorithm is also proposed on the bases of H.265/HEVC.The specific research content is as follows:This paper proposed an image coding algorithm based on measuring the perceptual distortion.Considering that the texture complexity and the region of interest will affect the perception of video quality,this paper designed subjective experiments for regions of interest and regions of non-interest under different texture complexity,and obtained the relationship between quantization parameters and subjective distortion levels.According to the experimental results,the perceptual distortion measure of the image is established.According to this measure,a perceptual distortion optimization method is proposed and applied to the HM16.5 software.Experimental results showed that compared to the HM16.5 original algorithm,the perceptual-distortion-based image coding algorithm proposed in this paper has an average improvement of 6.78% in the subjective quality of the reconstructed image at the same code rate.This paper also proposed a video coding algorithm based on measuring the perceptual distortion.Firstly,this paper proposed a description method of video motion features,and then designed subjective experiments on videos with similar spatial features and different motion features.According to subjective experiment we obtained the relationship between quantization parameters and subjective distortion levels.Combining the above experimental results and the proposed image perceptual distortion measure,this paper established a video perceptual distortion measure.A rate distortion optimization method based on the video perceptual distortion measure is further propsoed and applied to the HM16.5 software.Experimental results showed that compared to the HM16.5 original algorithm,the perceptual-distortion-based video coding algorithm proposed in this paper is able to increase the subjective quality of the reconstructed video by an average of 4.87% at the same code rate.