基于视觉颜色对比敏感度模型的动态视频压缩预处理算法

针对彩色视频压缩过程中压缩率不高或者视频质量损失过大的问题,提出一种基于视觉感知模型与色差的自适应视频压缩预处理算法。首先,在颜色对比敏感度模型中,人类视觉对高色度区域的色差敏感度较低,因此为不同色度区域分配不同的压缩权重；然后,基于帧之间的偏差图建立运动预测帧的空间运动模型,降低了高运动区域的残差；最终,使用动态的色调映射函数来控制视频的压缩等级,从而保证视频的视觉质量。基于多组视频进行实验,本算法成功地将标准的压缩软件提高了35%的压缩率,但同时可看出,本算法的性能依赖于具体的视频上下文。

Visual color contrast sensitivity model based dynamic preprocess schema of video compression

Aimed at the problem that in the color video compression process the compress rate is not high and the video quality loss is high, a visual color contrast sensitivity model based dynamic preprocess schema of video compression is proposed. Firstly, in the color contrast sensitivity model, human visual is not sensitive to high chroma region, so different compression weights are assigned to different chroma regions; then, the space motion model of motion prediction frame based on variance map of frame and frame, and the residual of the high motion region is reduced; lastly, adaptive tone mapping functions are used to control the compress rate of the video, so the visual quality of video could be high. Some experiments results show that the proposed algorithm got 37% compress gain than referred compression software, at the same time, the performance of the proposed algorithm depends on the context of the video.