Efficient Video Coding Optimization Using a Novel Perceptual Distortion Model

With the latest H.264/AVC standard, coding performance has made a tremendous progress. However, coding strategy is as important as coding standard which helps further improve the coding efficiency. People are eager to achieve less distortion with less bandwidth consumption. In most cases, fixed coding parameters are allocated to each unit regardless of the difference of the perceptual characteristics. Since videos are always received by human eyes, observation characteristics should be considered in coding strategies. Distortions should be reasonably arranged to fit the subjective sense. In this paper, firstly a new model between perceptual distortion and traditional distortion (like MSE) is established. Then the new model is used to optimize two critical modules in video coding framework, the rate-distortion optimization module and the quantization control module, in order to make the overall coding result suit for human observation. The simulation result shows that our proposed strategy is fit well for the H.264/AVC. The quality can be improved in both subjective and objective tests.