基于SSIM与ROI的视频质量评价方法

Video quality assessment based on SSIM and ROI

在客观结构相似视频质量评价模型SSIM基础上,提出了一种基于感兴趣区域(ROI)的视频质量评价方法,方法通过对帧内感兴趣区的提取,对帧内分块计算的SSIM值进行加权。该方法具备了SSIM算法的简单、高效等特性,同时又融合了人类感兴趣区域这一视觉特性,使得算法更好的反映了人的主观感受。在VQEG PhaseI测试数据集的实验结果表明,该方法的非线性回归后相关系数、斯皮尔曼相关系数、线外率等指标均优于传统的其他视频质量评价算法,有效地提高了视频客观评价与主观评价的相关度。

The paper presents a video quality assessment method based on the structural similarity index （SSIM） and the region of interest （ROI）. In the proposed method, the SSIM values for those intra blocks are calculated using different weighted values after the ROIs are extracted. The method has properties of simple and efficient as the same of the SSIM method. And it is more suitable for human visual system because of fusing the ROI. Experimental results show that the method can reflect people＇s subjective feelings in a better way and is better than other traditional methods in fitting M2 （Non-linear regression）, M3 （Spearman）, M4 （Outlier Ratio） of VQEG Phase I MOS.