一种基于互信息的图像质量评价改进方法

An Improved ImageQualityAssessment Method Based on Mutual Information

针对结构相似度方法在图像评测时基于平均加权策略的不足,基于信息论的思想,提出了一种改进的基于互信息的图像质量评价方法。该方法采用高斯尺度混合模型对图像进行建模,将参考图像与失真图像的互信息作为质量评价函数的视觉权重,设计了一种基于互信息和人类视觉特性的质量评价测度函数。该方法充分利用了人类视觉的感知特性以及图像内容特性,从整体上更好地反映了人的主观感受。仿真结果验证了该改进方法的有效性,并且在性能上优于传统的方法。

According to the shortcoming of structural similarity method based on weighted average which for evaluating the quality of image,an improved image quality assessment method based on mutual information is proposed based on the view of information theory.This method uses Gaussian scale Mixture to model the image based on structural similarity method,thus apply the mutual information between the reference image and distortion image as the visual weighting of the function of quality assessment.Finally,a function of quality assessment is designed based on mutual information and human vision characteristic.The methods takes advantage of the perceptual characteristics and image content characteristics of human vision and reflect the people＇s subjective feelings.Simulation results show the effectiveness of the improved method and show that the proposed method is better than the traditional methods.