文献类型：期刊文献

中文题名：基于统计及结构稀疏特性的图像修复算法

英文题名：IMAGE INPAINTING ALGORITHM BASED ON STATISTICS AND STRUCTURE SPARSITY CHARACTER

作者：于文静[1];毕东旭[1];颜学峰[1];林家骏[1]

机构：[1]华东理工大学信息学院,上海200237

年份：2016

卷号：33

期号：3

起止页码：214

中文期刊名：计算机应用与软件

外文期刊名：Computer Applications and Software

收录：CSTPCD;;CSCD:【CSCD_E2015_2016】；

语种：中文

中文关键词：图像修复;灰度均值直方图;结构稀疏

外文关键词：Image inpainting; Gray-mean histogram; Structure sparsity;

摘要：目前普遍使用的基于等照度线的优先权图像修复算法,不能快速准确地确定待修复图像的结构位置。提出利用破损区域边缘图像块的灰度均值直方图,快速准确定位结构点的位置,优先修复结构区域后,实现纹理区域修复,获得很好的修复结果。该算法与传统算法相比,不必计算每个边缘图像块的优先级,直接利用统计信息获得结构点的位置,比传统算法更准确获取结构点位置。算法的计算效率和修复效果都有了很大提升。
Current popular image inpainting algorithm is based on isophote priority,which cannot determine the structure location of the image to be restored fast and accurately. Our algorithm suggests to use gray-mean histogram of the patch on rim of the deteriorated region to rapidly position the location of structure points,and after a prior inpainting on structure region,the filling-in of texture region will be implemented. This achieves very good restoring result. Comparing our algorithm with the classical algorithm,it does not need to calculate the priorities of every edge patch but directly makes use of statistics information to obtain the location of structure points,and is more accurate than acquired by traditional algorithm. Both the computation efficiency and inpainting effect of this algorithm are greatly improved.