结合局部特征的无参考彩色图像质量评价

由于传统的无参考彩色图像质量评价方法与人眼感知结果的一致性较差,本文提出了一种全面利用待评价图像的色度、锐利度和对比度的无参考彩色图像质量客观评价方法。分析了彩色图像锐利度的局部特征,提出了一种新的彩色图像锐利度测量模型。基于对比度的局部特征和Buchsbaum曲线特征,建立了新的彩色图像对比度测量模型。最后,通过线性组合色度测量模型、锐利度测量模型和对比度测量模型,构建了无参考彩色图像质量评价函数。利用TID2013数据库中的3类退化图像(高斯模糊图像、对比度改变图像和噪声图像)验证了本文提出的锐利度测量模型、对比度测量模型和无参考彩色图像质量评价函数的性能。结果表明,本文提出的锐利度测量模型和对比度测量模型的性能均优于传统的锐利度和对比度计算模型。提出的无参考彩色图像质量评价函数的Spearman秩相关系数(SROCC)为0.904,Kendall秩相关系数(PROCC)为0.865,Pearson线性相关系数(PLCC)为0.922,亦均优于传统方法。

No-reference color image quality assessment based on local features

For the poor consistency of traditional no-reference color image quality assessment methods and human visual perceptive results, a new no reference color image quality assessment method was proposed based on the colorfulness, sharpness and contrast of images. A new sharpness measuring model for a color image was proposed based on the local feature of sharpness. Then, a novel contrast measuring model for the color image was established based on the local feature of contrast and the feature of Buchsbaum curve. Finally, a novel no-reference color image quality assessment function was constructed based on the linear combination of the colorfulness measuring model, sharpness measuring model and the contrast measuring model. The performance of the sharpness measuring model, contrast measuring model and the no-reference color image quality assessment function was verified by three kinds of degraded images(Gaussian blurred image, contrast changed image and noise image). The experiment results indicate that comparing the traditional methods, the performance of the proposed sharpness measuring model and the contrast measuring model is better than that of the traditional ones. The Spearman Rank Order Correlation Coefficient(SROCC), Kendall Rank-Order Correlation Coefficient(KROCC), and the Pearson Linear Correlation Coefficient(PLCC) of the proposed color image quality assessment function are 0.904, 0.865 and 0.922, respectively, which has better consistency as compared with those of the traditional methods.