基于2D视觉注意模型的全参考图像质量评价方法

为进一步提高视频图像质量评价主客观的一致性, 提出一种基于2D视觉注意模型的全参考质量评价方法。将2D视觉注意模型与PSNR(Peak Signal to Noise Ratio)、 SSIM（Structural Similarity Index Method）、 GSSIM(Gradient based Structural Similarity Index Method)评价方法相结合, 进行视频图像质量评价。实验结果表明, 该方法对图像质量评价较原PSNR、SSIM、GSSIM方法更有效； 在LIVE数据库上得到的Pearson、Spearman相关系数分别为0.903 5和0.903 1； 较其他评价方法原理简单, 计算复杂度较低, 且性能较优, 具有很好的评价效果。

Image Quality Assessment Method with Full Reference Based on 2D Visual Attention Model

In order to further improve the consistency of the objective and subjective video image quality assessment, we present a quality evaluation method with full reference based on 2D visual attention model, which combines the 2D model with PSNR （Peak Signal to Noise Ratio）, SSIM （Structural Similarity Index Method） and GSSIM （Gradient based Structural Similarity Index Method） evaluation methods for quality evaluation. The experimental results show that the proposed methods of image quality assessment are more effective than PSNR, SSIM and GSSIM methods. Tested on LIVE database with its Pearson, Spearman Correlation equals to 0. 903 5 and 0. 903 1. Comparing with other evaluation methods, it is simpler in principle, lower in calculation complex and has better performance and better evaluation results.