近年來,應用人眼視覺特性於視覺編碼的研究已經日益受到重視,然而在一般的視訊編碼標準中仍然很少考慮到人眼視覺特性對視訊的影響去做壓縮或編碼。在本篇論文中,將人眼視覺的特性加入H.264/AVC的編碼器中,對於各種壓縮的模式,提供一個選擇器,選擇器中主要是以視覺上的最小 rate-distortion cost為參考依據,從各種inter/intra modes中選擇出一個最好的壓縮模式,使得新的視訊和H.264相較之下,可以得到在視覺上更好的影像品質。另外,經由加入判斷式來分析skip壓縮模式對視覺影像品質的影響,節省在視覺上擁有可容許視覺失真的macro-block的bits,分配給有較大視覺失真值的macro-block使用,在相同的bit rate下就能夠把視訊重要的部分表現得更加清楚。根據我們的實驗數據顯示,利用既有的視覺標準來作為判斷,可以比H.264有更好的視覺品質。
Video coding based on human visual system (HVS) has become an important subject in the area of video communication. However, the HVS is rarely used in standard codec. This paper focus on how to enhance the perceptual quality of H.264/AVC codec by means of improving the mode selection algorithm based on Human visual model. In the proposed mode selection algorithm, the perceptual minimum rate-distortion cost criterion would choose the best mode from various inter/intra prediction modes to optimize video quality on macro-block basis. In comparison with H.264 standard, our approach can choose the perceptually optimized compression mode for coding. By introducing threshold to adjust compression modes, the quantity of bits in each macro-block can be optimized so that important parts of frames can be displayed under the same bit rate. The simulation results show that, according to visual perception criterion, our proposed algorithm enhances video qualities compared to conventional H.264 standard.