Researches On Object-based Perceptual Video Coding

The rapid development of information technology makes the traditional information processing methods and performance means difficult to meet the needs of the community. The new video application needs a higher demand for video coding performance and stability. On the basis of the traditional coding technology, people focus more attention on visual perception.The content of this thesis is object-based video perceptual coding technology with the introduction of visual perception features to improve the coding efficiency. The specific research contents include the following aspects.(1)The human visual system, stereo imaging principle and the display technology of three-dimensional video are explored. A detailed analysis of the mechanism of human visual system is given firstly. Then the human visual characteristics are integrated into the video coding technology, aiming to improve the efficiency of video compression coding. At last, a summary of the existing common stereo display technology is introduced.(2)A variety of single channel and multi-channel object extraction algorithms are studied. In the process of encoding, different object and region will get different attention from human eyes, so they share different coding features in theory. A optimal coding strategy must be selected for the regions of interest to ensure the perception quality, whereas, less bits can be distributed to other regions without affecting the overall perception quality. Extract the salient objects in the video,therefore, not only can speed encoding, but also can save bits to a certain extent.(3)A new video coding algorithm is proposed based on the saliency object. Research shows that eyes attention abides by certain rules during you watching the video, noticing a few objects and regions that cause a strong stimulus to human eyes at first and transferring to other objects and regions that cause a weak stimulus. In saliency based video coding, the salient parts of the each frame are encoded with higher quality than non salient areas to improve the signal to noise ratio and get higher compression ratio. In this thesis, the quantization control in high efficiency video coding(HEVC) is incorporated with the visual saliency map to reduce the bit rate by adaptively changing the quantization parameter. Experimental results show that the proposed method can save the bit rate with improving the perceptual quality of regions-of-interest.