基于区域对比度和SSIM的图像质量评价方法

在诸多图像质量评价方法中，结构相似度(SSIM)算法简单高效，准确性较高，但SSIM模型不能很好地评价存在局部失真和交叉失真类型的图像。针对SSIM算法对图像不同区域平等对待的不足并考虑了时域人眼视觉特性，提出一种改进的基于区域对比度和结构相似度(RCSSIM)的图像质量评价方法。该算法将图像区域灰度信息对比度与SSIM算法融合，加权归一为参考图像与失真图像的对比度结构相似度值，以其评价图像质量。在LIVE图像数据库上的实验结果表明，与SSIM算法相比，RCSSIM评价结果的皮尔逊线性相关系数提高约0.015，均方根误差减小约0.55，更接近于人眼主观测试结果，具有更好的评价性能。

Image quality assessment method based on regional contrast and structural similarity

Among numerous image quality assessment(IQA) methods, the structural similarity (SSIM) algorithm is simple, high efficient and accurate. However, it often does not work well when there is regional distortion or cross distortion in the image. To deal with the problem that SSIM algorithm treats the different regions of the image identically, we took human visual characteristics in spatial domain into consideration and put forward an improved IQA method based on regional contrast and structural similarity(RCSSIM). The new algorithm combines regional contrast with structural similarity, weighs and normalizes the original SSIM index to a regional contrast structural similarity metric between the reference image and the distortion image to assess the image quality. The experiment results on LIVE image database show that the Pearson linear correlation coefficient(PLCC) of the new algorithm increases by about 0.015 and the root-mean-square error decreases by about 0.55 compared with the SSIM algorithm. It indicates that the evaluation result of RCSSIM algorithm is more consistent with human visual system(HVS) characteristics and is more effective than the SSIM algorithm.