基于颜色与结构失真的色域映射图像无参考质量评价算法

No-reference quality evaluation algorithm for gamut mapping images based oncolor and structural distortions

为了预测色域映射图像客观质量,通过分析不同色域映射算法的映射原理发现色域映射图像中主要存在颜色失真与结构失真的情况。基于此,提出了一种基于颜色与结构失真的色域映射图像无参考质量评价算法。在颜色失真方面,计算色调异常率和图像R、G、B三个分量的统计分布与理想均匀分布之间的相对熵;在结构退化方面,提取图像的信息熵与四阶矩,并对图像亮度与饱和度进行统计建模,提取参数特征。随后,将以上提取的数据作为质量感知特征与图像的主观分数值输入后向传播神经网络进行回归训练得到针对色域映射图像的质量评价模型。最后,在三个公开的色域映射图像数据库上进行性能验证。实验结果表明,该算法在预测色域映射图像质量方面优于现有的无参考算法。

In order to objectively predict the quality of gamut mapping images,this paper analyzed the mapping principles of different gamut mapping algorithms and found that there were mainly color and structural distortions in gamut mapping images.Based on this,this paper presented a no-reference quality metric for gamut mapping images based on color and structural distortions.For color distortion,this paper calculated the rate of abnormal hue and the Kullback-Leibler divergence between the statistical distribution of the three components of the image(e.g.,R,G,and B)and the ideal uniform distribution.In terms of structural distortion,it extracted the entropy and the fourth-order moments,and extracted statistical features in brightness and saturation components.Subsequently,combined with the subjective scores and extracted features,it used the back propagation neural network(BPNN)to train the quality prediction model.Finally,this paper employed the model to evaluate the quality of gamut mapping images.Extensive experiments conducted on three gamut mapping databases prove the proposed method is superior to the existing quality evaluation models in evaluating the quality of gamut mapping images.