基于L_(0)矩阵范数正则化的自然图像去反光算法  

Highlight removal algorithm for real-world images based on L_(0)matrix norm regularization

作　　者：丁凤 夏又生 DING Feng;XIA Yousheng(College of Mathematics and Statistics,Fuzhou University,Fuzhou,Fujian 350108,China;College of Artificial Intelligence,Nanjing University of Information Science and Technology,Nanjing,Jiangsu 210044,China)

机构地区：[1]福州大学数学与统计学院,福建福州350108 [2]南京信息工程大学人工智能学院,江苏南京210044

出　　处：《福州大学学报（自然科学版）》2022年第6期729-736,共8页Journal of Fuzhou University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61473330)。

摘　　要：提出一种基于L_(0)范数正则化的自然图像去反光算法.首先,根据自然反光图像的两个特征构建基于L_(0)范数的正则优化模型,保证漫反射图像系数矩阵的稀疏性、低秩性和反光区域漫反射分量的有效恢复.其次,利用增广拉格朗日技术,导出求解L_(0)范数正则优化模型的算法.最后,通过与相关的图像去反光算法对比,证实本图像去反光算法在均方误差和结构相似度上均优于其他去反光算法,使其生成图像在保留更多纹理细节信息的同时,可以有效去除图像反光.A real-world image specular highlight removal algorithm based on L_(0)norm regularization is proposed.Firstly,according to the two characteristics of natural reflective image,a regular optimization model based on L_(0)norm is constructed to ensure the sparsity,low-rank of the diffuse image coefficient matrix and the effective recovery of the diffuse reflection components in the highlight region.Secondly,using the augmented Lagrange technique,an algorithm is derived to solve the L_(0)norm regular optimization model.Finally,by comparing with related image reflection removal algorithms,the experiment proves that the proposed image reflection removal algorithm is superior to other image reflection removal algorithms in terms of mean square error and structural similarity.The generated image can effectively remove image reflection while more texture details are retained.

关 键 词：图像反光去除 L_(0)范数正则化 矩阵变量优化 

分 类 号：TN911.73[电子电信—通信与信息系统] TP301.6[电子电信—信息与通信工程]