一种新的结构特征全参考图像质量评价方法

A Novel Structural Feature Similarity Index for Full Reference Image Quality Assessment

提出一种基于单演信号理论提取结构特征的全参考图像质量评价方法。方法首先利用单演信号理论将图像分解为幅值、方向和相位3个特征分量,构造单演相位一致映射图和特征分量相似度比较函数;然后对相位一致映射图进行Riesz变换,其变换后的一阶、二阶系数作为结构特征相似度比较函数,最后将单演信号的特征分量和结构特征相似度比较函数归一化加权计算得出最终的图像质量评价值。实验结果表明,由于单演相位一致具有较好的抗噪声能力,特征分量和结构特征相似度更加全面考虑到对于人眼感知图像结构的重要性,因此评价结果与图像主观质量评价具有较好的一致性。

A novel feature similarity( RMFSIM) index for full reference IQA based on monogenic signal theory is proposed. First,image is decomposed into three orthogonal components of local amplitude,local phase and local orientation using monogenic signal theory. Comparison functions which are similarity functions of monogenic phase congruency map and feature components are constructed. Then,the coefficients maps of the 1st-order and the 2ndorder Riesz transforms of monogenic phase congruency map act as structure similarity comparison function. Finally,the final image quality evaluation is obtained by calculating and normalizing comparison functions weighted. Experimental results demonstrate that since the importance of structure magnitude and orientation for the human eyes is more comprehensive consideration the proposed similarity index is highly consistent with human subjective evaluations and achieves good performance in terms of prediction monotonicity and accuracy.