基于改进样本块匹配准则的数字图像修复方法  

A digital image restoration method based on improved sample block matching criterion

作　　者：张园园[1] ZHANG Yuanyuan(Anhui Technical College of Industry and Economy,Hefei 235100,China)

机构地区：[1]安徽工业经济职业技术学院计算机与艺术学院,安徽合肥235100

出　　处：《商丘师范学院学报》2022年第3期15-18,共4页Journal of Shangqiu Normal University

摘　　要：为解决以往数字图像修复方法存在的对图像潜在亮度敏感度高,修复后的图像存在变形差异的问题,提出基于改进样板块匹配准则的数字图像修复方法.采用Criminisi算法更好完成样板块配准,获取最佳目标匹配块;采用基于IG-GN算法的图像修复方法利用最佳目标匹配块,通过最小化ZNSSD相关准则实现数字图像修复,且修复过程可显著降低图像块之间的差异,保证图像块之间的纹理及结构特征.实验得到:该方法修复后的数字图像没有明显模糊区域和边界纹路,使修复后的图像更接近原图或者达到原图,修复效果好.In order to solve the problem of high sensitivity to the potential brightness of the image in the previous digital image restoration methods, and the difference in the deformation of the repaired image, a digital image restoration method based on the improved template block matching criterion is proposed.Criminisi algorithm is used to better complete the template fast registration and obtain the best target matching block;ig-gn algorithm based image restoration method uses the best target matching block to realize the digital image restoration by minimizing the znssd correlation criteria, and the restoration process can significantly reduce the differences between image blocks and ensure the texture and structural features between image blocks.The experimental results show that the restored digital image has no obvious fuzzy areas and boundary lines, which makes the restored image closer to or reach the original image, and the effect of restoration is good.

关 键 词：改进样本块 匹配准则 数字图像 修复 IC-GN算法 图像纹理 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]