非局部多尺度分数阶微分图像增强算法研究  

Research on Non-local Multi-scale Fractional Differential Image Enhancement Algorithm

作　　者：黄果[1] 许黎[2,3] 陈庆利 蒲亦非[3] HUANG Guo;XU Li;CHEN Qingli;Pu Yifei(Sichuan Province University Key Laboratory of Internet Natural Language Intelligent Processing,Le shan Normal University,Leshan 614000,China;School of Physics and Electronics,Le shan Normal University,Leshan 614000,China;School of Computer Science,Sichuan University,Chengdu 610064,China)

机构地区：[1]乐山师范学院互联网自然语言智能处理四川省高等学校重点实验室,乐山614000 [2]乐山师范学院物理与电子工程学院,乐山614000 [3]四川大学计算机科学学院,成都610064

出　　处：《电子与信息学报》2019年第12期2972-2979,共8页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(61201438);四川省科技厅应用基础项目(2016JY0238);四川省教育厅重点项目(18ZA0235);四川省教育厅一般项目(18ZB0268,18ZB0266);乐山师范学院科研项目(JG2018-1-04)~~

摘　　要：为了更好增强图像中的有用信息,改善图像视觉效果,该文提出了一种基于非局部多尺度分数阶微分图像增强算子(NMFD)。该算子首先将图像分成若干块子图像,计算每一块子图像的边缘强度系数、熵值和粗糙度等细节特征,将得到的特征数据在全局图像范围进行统一尺度的归一化,然后对这些归一化的数据进行加权求和作为图像的非局部特征值,最后利用指数函数建立图像细节特征和分数阶微分算子阶次之间的非线性量化关系,在不同的图像子块区域,确定不同尺度的分数阶微分阶次,实现图像的非局部多尺度增强。In order to enhance the useful information in the image and improve the visual effect of the image,a Non-local Multi-scale Fractional Differential(NMFD)image enhancement operator is proposed.The operator divides the image into several sub-images and calculates the edge intensity coefficient,entropy value and roughness of each sub-image,and the obtained feature data are normalized in a unified scale in the global image range.Then,the normalized data are weighted to be the non-local eigenvalues of the image.Finally,an exponential function is used to establish the non-linear quantization relationship between image detail features and the value of fractional order.Thus,the fractional order of different scales can be determined in different image sub-block regions,so that the non-local multi-scale image enhancement model is realized.

关 键 词：图像增强 非局部多尺度分数阶微分算子 图像熵值 图像对比度 

分 类 号：TN391[电子电信—物理电子学]