基于RDO和SSIM的AVS帧内模式选择算法

Improvements for AVS Intra Mode Selection Based on RDO and SSIM

AVS帧内模式选择算法中,常采用率失真优化(RDO)方法作为最后模式选择的代价函数来对帧内的5种预测模式进行选择。虽然该方法可以选出相较绝对误差和(SAD)方法更优的预测模式,但是仍不能很好地满足人眼视觉系统的主观性。结构相似度(SSIM)作为一种新的图像质量评价方法,不仅算法简单且包含符合人眼视觉感知的结构信息,故更优于传统的PSNR等评价方法。对原有帧内模式选择算法进行改进,将SSIM和RDO结合在一起共同作为帧内模式选择的代价函数。实验结果表明,该新的模式选择算法获得了更好的编码质量和更高的编码效率。

In the AVS intra mode selection algorithm ,rate distortion optimization （RDO） method is used as a last mode selected in the cost function of intra prediction mode selection. This approach can elect better prediction mode than with absolute error（SAD） method,but this method still can not meet the subjectivity of the human visual system. Structural similarity（SSIM） as a new image quality evaluation method, not only is simple, but also contains the structural information of human visual perception. So it is better than the traditional PSNR evaluation method. In this paper,the old intra mode selection algorithm is improved ,SSIM combined RDO together as the cost function in intra mode selection. Experimental results show that this new mode selection algorithm get a better coding quality and a higher coding efficiency.