摘要
图像修复是一种去除图像中多余物体并采用合理的纹理去填充待修复区域的技术。针对Criminisi等人提出的基于样本的图像修复算法,根据目标块中的结构信息与中心点梯度值的相关性,提出了一种新的优先权函数设计方法,以保持待修复区域的强连续性。为加快匹配效率,在确保要获取的匹配块在搜索范围内的同时,采用局部搜索匹配块的方法,相比Criminisi原算法中通过全局搜索得到匹配块的方式,速率更快。实验结果表明,该算法能获得更佳的修复效果。
Image inpainting is technique for removing useless objects and filling the unknown regions with reasonable texture. Based on the sample-based image restoration proposed by Criminisi.According to the correlation between the structure information in the target block and the gradient of center point.This paper presents a new method of the priority function design to maintain the strong continuity of the unknown regions.In order to speed up the matching efficiency and ensure the matching block is in the search range,use the local block matching.Compared with the original algorithm of obtaining the matching block by global searching in Criminisi,Our method runs faster.
出处
《工业控制计算机》
2014年第7期127-128,131,共3页
Industrial Control Computer
