自适应匹配窗口的Criminisi图像修复算法  

Criminisi Image Restoration Algorithm based on Adaptive Matching Window

作　　者：周彩月 周崇波[1] 吴冬梅 孙琳[1] ZHOU Caiyue;ZHOU Chongbo;WU Dongmei;SUN Lin(Qufu Normal University,Qufu Shandong 273165,China)

机构地区：[1]曲阜师范大学,山东曲阜273165

出　　处：《通信技术》2021年第5期1087-1094,共8页Communications Technology

基　　金：山东省自然科学基金(No.ZR2017MD019);山东省高校科技计划项目(No.J15LN08)。

摘　　要：针对Criminisi算法在图像修复中存在匹配错误、匹配标准不准确等问题,提出了一种改进的Criminisi修复算法。首先利用边缘特征和局部亮度特征,改进优先权模型以增进对边缘和纹理的辨别;其次利用已修复点的统计信息,以自适应的方式确定搜索匹配窗口大小;最后将基于像素和统计的度量结合起来,并引入互信息作为新的匹配准则。实验结果表明,提出的改进修复算法优于其他算法,能更有效恢复大面积缺失图像,处纹理细节保持效果较好。Aiming at the problems of Criminisi algorithm in image inpainting,such as matching error and inaccurate matching standards,an improved Criminisi inpainting algorithm is proposed.Firstly,edge features and local brightness features are used to improve the discrimination of edges and textures;Then,the statistical information of the repaired points is used to determine the size of the search matching window in an adaptive manner;Finally,the measurement based on pixel and statistics is combined,and mutual information is introduced as a new matching criterion.Experimental results indicate that the proposed improved algorithm is better than other algorithms,and it can more effectively restore large-area missing images,and maintain better texture details.

关 键 词：Criminisi算法 图像修复 边缘特征 自适应匹配窗口 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]