一种针对高清图像的JND模型改进方法

Improved Algorithm for HD Picture Quality Assessment Based on JND Model

基于全参考的JND(刚辨差)模型是视频图像质量客观评价算法中的一个经典算法。对该算法的重要步骤和主要参数进行深入分析,并结合高清视频图像的特点,在通道分解、参数选取和屏蔽效应等方面提出改进思路并建立用于高清视频的改进型JND模型。实验结果表明,在评价高清视频图像质量时,该模型可获得与主观评价结果更为一致的客观测量结果。

The JND model is one of classic metrics in objective perceptual video quality research. In this paper, thorough analy- sis of this model is given and an improved algorithm for HD （High Definition） video is proposed to obtain better objective-to- subjective correlation. This improved model is optimized from the following aspects including multi-channel decomposition, parame- ter selection and masking effects. Experimental resuhs show that our model can achieve good performance for HD video quality prediction.