结合HVS及SSIM的无参考模糊图像评价方法  

No-reference blur image quality metric combining HVS with SSIM

作　　者：袁万立[1] 李朝锋[1] 

机构地区：[1]江南大学物联网工程学院,江苏无锡214122

出　　处：《计算机工程与应用》2013年第1期210-212,221,共4页Computer Engineering and Applications

基　　金：国家自然科学基金(No.61170120);中央高校基本科研专项资助(No.JUSRP20915);江苏省自然科学基金(No.BK2011147)

摘　　要：针对传统无参考模糊图像质量评价算法未考虑人类视觉系统在模糊检测时的有限性这一缺点,提出了一种结合人类视觉特性及结构相似度的无参考图像质量评价方法。该方法向原始输入图像增加低通滤波器生成重新模糊图像;之后创建一个以原始输入图像的边界点为中心的强边界像素块集合;分别计算集合中像素块与对应的重新模糊图像中像素块间的结构相似度指数;所得的全部结构相似度指数的均值作为最终的原始输入图像的模糊值。基于LIVE模糊图像数据库的实验结果表明,该方法有较好的性能,其计算结果与主观统计值的线性相关系数及斯皮尔曼等级相关系数分别为0.9298及0.9318。The traditional no-reference blur metrics haven＇ t considered the limitation of the Human Visual System （HVS） in blur detection. A novel no-reference blur metric combining characteristics of HVS with Structure Similarity（SSIM） is proposed. In the new method, a new re-blurred image is produced by convoluting the original image with a low pass filter. Then a collection of strong edge block around edge point detected by the Sobel operator is created. For each pixel block, the SSIM index of it and corresponding pixel block in re-blurred image is calculated. At last, the new blur metric as the average of the SSIM indices is taken. The experimental results on the LIVE blur database demonstrate that the method can obtain good performances. The linear corre-lation coefficient and the Spearman rank correlation coefficient between the results of the proposed method and the subjective quality measurements are 0.929 8 and 0.931 8 respectively.

关 键 词：人类视觉系统 结构相似度 模糊估计 无参考 图像质量评价 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]