摘要
基于纹理的图像修复算法对于修复破损区域比较大的图像效果较好,但该算法对于含有结构信息的图像修复效果很差.通过新的优先项的计算、平均值补偿及增加惩罚项提高传统的基于样本的图像修复算法的修复效果,结合图像中常出现的直线和曲线结构特征,提出了基于样本和结构信息的大范围图像修复算法.实验表明,该算法易于实现,修复结果能达到令人满意的效果,具有较高的实用价值.
Image inpainting algorithms prefer to use the texture-based method to inpaint large scale missing regions, but the result is undesirable when the image contains structural information. Through introducing new priority value, mean value complement and penalty term in the traditional exemplar based inpainting method to improve the performance of it. After combining the linear and curve structural features, we propose an exemplar and structure based inpainting algorithm for large scale missing regions in this paper. The new algorihm is easy to implement and the inpainted experimental results compared with some inpainting algorithms are plausible.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2012年第8期1509-1514,共6页
Acta Electronica Sinica
基金
国家自然科学基金(No.50805031)
深圳市科技计划项目(No.JC201005260161A)
深圳市三大产业专项资金(No.JC201104210015A)
深圳南山科技计划项目(No.201002)
关键词
图像修复
纹理
样本
结构信息
image inpainting
texture
exemplar based method
structural information
