基于局部结构张量的SSIM图像质量评价  

Structural Similarity for Image Quality Assessment Based on Local Structure Tensor

作　　者：罗学刚[1,2] 吕俊瑞[2] 王华军[1] 

机构地区：[1]成都理工大学地球探测与信息技术教育部重点实验室,四川成都610059 [2]攀枝花学院数学与计算机学院,四川攀枝花617000

出　　处：《微电子学与计算机》2014年第9期109-112,共4页Microelectronics & Computer

基　　金：国家自然科学基金青年科学基金项目(61202195/F020502)

摘　　要：针对结构相似度(SSIM)评价模糊图像存在的局限性,提出一种基于局部结构张量的结构相似性图像质量评价方法(ST-SSIM).方法首先利用结构张量提取图像几何结构强度和方向角度信息,然后用局部结构方向相似度比较函数和结构强度相似度比较函数替换SSIM中原有的结构比较函数,最后在图像质量映射图上以参考图像的结构强度函数加权计算得出最终的图像质量评价值.实验结果表明,由于LST-SSIM更加全面考虑到对于人眼感知结构强度和方向的重要性,因此评价结果比SSIM更加符合人眼视觉感知特性.Aiming at the drawback of SSIM on blurry images,an improved Structural Similarity method for image quality assessment based on structure tensor is presented.Firstly,geometrical structure magnitude and orientation angle are extracted by local structure tensor.Secondly,a comparison function which is a structural orientation angle function and a structural magnitude function replaces the structure comparison function in the original SSIM.Finally,we use structure magnitude as a weighting function to derive the local quality map to obtain the final image quality evaluation value.The results show that since the importance of structure magnitude and orientation for the human eyes is more comprehensive consideration in perceiving structural information,the ST-SSIM is more accordant with the characteristics of human visual perception than SSIM.

关 键 词：图像质量评价 结构信息 结构相似度 局部结构张量 人眼视觉系统 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]