基于Contourlet域梯度结构相似度的无参考图像质量评价

Contourlet变换已经成为多尺度、多方向图像处理领域的研究热点,文章将其引入图像质量评价,提出了一种基于Contourlet变换的梯度结构相似度无参考图像质量评价方法（NRCT-GSSIM）,将待评价图像通过高斯低通滤波器构造参考图像,然后在Contourlet变换域中计算图像频率子带、方向子带分别加权的梯度结构相似度,实现对图像的质量评价.该方法无需标准参考图像,充分考虑了人类视觉特性.实验结果表明,NRCT-GSSIM取得了与峰值信噪比（PNSR）、结构相似度（SSIM）和梯度结构相似度（GSSIM）等有参考图像质量评价相一致的评价结果,且较之PSNR、SSIM有更高的准确性.

No-Reference Image Quality Assessment Based on Gradient Structural Similarity in Contourlet Domain

The contourlet transform has drawn tremendous attention in the research of multi scale and multi direction image processing. This paper introduces it into image quality assessment and provides a new image assessment method based on the contourlet transform and gradient structural similarity without standard images, which is called NRCT GSSIM. This method constructs a reference image by a Gaussian low pass filter and assesses the image quality by computing weighting gradient structural similarity either in frequency sub bands or in direction sub bands. The new method considers features of the human vision system and it does not need the standard reference image. Simulated results show that NRCT GSSIM is well in accordance with quality assessment results of some full reference methods such as the peak signal to noise ratio , structural similarity and gradient structural similarity indexes and the computed result of NRCT GSSIM is more accurate than the ones of PSNR and SSIM.