摘要
当前的图像修复算法都是利用非连续边缘的已知块信息来完成损坏区域的填充,造成图像模糊与视觉不连通;且修复路径都是随机确定,使其成本较高。对此,提出了拓扑梯度耦合多重最小路径快速行军的连续轮廓图像修复优化算法。引入拓扑梯度,检测出缺失区域的边缘轮廓;定义关键点择取规则,提取图像损坏区域的关键点,嵌入权重因子,建立权重距离函数,计算最小修补路径成本,并设计多重最小路径快速行军机制,提取出连续边缘,完成损坏区域填充。仿真结果显示,与其他图像修复算法相比,本文算法可检测出损坏区域的连续边缘轮廓;且该算法具有更好的修复视觉与效率。
The blurring and block effects of reconstruction image are induced by using the known block information with discontinuous edges to fill the damage area.So the image inpainting algorithm of connected contours based on topological gradient and multiple minimal paths fast marching methods is proposed.The edge contour in damage area is detected by introducing topological gradient,then the minimiing inpainting path is determined and multiple minimal paths fast marching method is designed by building the weighted distance function coupled the weighted operator and choosing the key points of damage area,and also the image inpainting feedback mechanism is contrasted based on structural similarity and distortion to optimize the inpainting image.Simulation results show that compared with other image inpainting algorithms,this algorithm can detect continuous edge profile of the damaged area,and has a better inpainting effect and efficiency.
出处
《测控技术》
CSCD
2015年第9期152-156,共5页
Measurement & Control Technology
基金
国家自然科学基金项目(11171096)
关键词
拓扑梯度
最小路径
快速行军
连续边缘轮廓
权重距离函数
图像修复
topological gradient
minimal path
fast marching
continuous edge contour
weighted distance function
image inpainting
