一种改进的快速图像修复算法

郭犇, 肖创柏, 孙翘楚, 刘璐

北京工业大学计算机学院



摘要:

目前基于结构的图像修复算法中,基于快速行进的图像修复算法能够简单快速的修复数字图像中的破损区域,但是修复效果一般,特别是边缘区域的保持效果较差.在快速行进算法中引入了梯度权函数和距离权函数,通过对邻近点的加权计算进行排序,然后按权值大小对破损区域进行逐步修复,并利用梯度排序对边缘进行保持.实验结果表明该算法修复效果要优于原快速行进算法,但在速度方面却难以保持.引入梯度和距离权函数可以使修复效果更好.

关键词:  图像修复  快速修复  结构模型

DOI：10.11918/j.issn.0367-6234.2011.05.023

分类号:TP391.41

基金项目:国家高技术研究发展计划基金资助项目(2009AA012437);北京市自然科学基金资助(4072004,4092006)



An improved fast image inpainting method

GUO Ben, XIAO Chuang-bai, SUN Qiao-chu, LIU Lu

College of Computer Science and Technology,Beijing University of Technology,100081 Beijing,China

Abstract:

In the current structure-based image inpainting algorithm,the Fast Marching Method can simply and rapidly repair the damaged region in the digital image.However,the effect is not very good,especially poor to preserve the edge information.By introducing the gradient function and the distance weight function,an improved algorithm has been proposed,which sorts the neighboring points with weight to repair the damaged region and uses the gradient-sorting estimation to preserve the edge.The results of experiments have indicated that the improved one is better.The gradient function and the distance weight function can make results more better.

Key words:  image inpainting  Fast Marching Method  structure model