结合HVS和相似特征的图像质量评估方法

Improved image quality assessment model incorporating HVS and FSIM

图像质量的客观评价方法研究在实现图像质量评价仪器化的过程中起到决定性的作用。在分析最新全参考图像质量评价算法:特征相似法(feature similarity,FSIM)的基础上,利用对比敏感度函数(contrast sensitivity function,CSF)算子以及离散余弦变换(discrete cosine transform,DCT)域的对比度掩盖效应,提出了一种改进的FSIM图像质量评价方法。该方法具有FSIM算法简单、高效等特性,同时又充分体现人眼视觉特性,更好地反映了人的主观感受。LIVE(laboratory for image and video engi-neering)测试数据集的实验结果证明,该方法在非线性回归后相关系数、斯皮尔曼相关系数、线外率等指标方面均优于传统的其他图像质量评价算法。

The research on objective image quality metric plays a crucial role in the instrumentation of image quality assessment. In this paper, a new full-reference metric for image quality assessment is proposed, which is based on the recent feature similarity （FSIM） index and incorporates proper human visual system （HVS） characteristics. This method improves FSIM using the contrast sensitivity function（CSF） operator and the contrast masking operator in DCT （discrete cosine transform） domain. To test the performance of the proposed metric, we have carried out exper- iments on LIVE （laboratory for image and video engineering） database. Experimental results demonstrate that the improved metric can achieve higher consistency with the subjective evaluation than FSIM and other relevant state-of- the-art image quality assessment metrics.