现代电子技术2017，Vol.40Issue(11)：53-57,5.DOI:10.16652/j.issn.1004-373x.2017.11.014
Criminisi图像修复算法的优化
Optimization of Criminisi algorithm for image inpainting
摘要
Abstract
Since the Criminisi algorithm has the problems of structure fracture and wrong matching while making image inpainting, some new improvement thoughts are proposed on the basis of the original algorithm,with which the calculation method of the priority function is improved,and the priority function is represented as the weighted sum of the data item,confidence coefficient item and neighborhood correlation term to ensure the information continuity of the image structure. An algorithm of variable-size exemplar block was designed to enhance the local harmony and boundary smoothness. The similarity measure method for templates was im-proved,and integrated with color histogram to improve the accuracy of template matching. The experimental results show that the method can make up the deficiencies of Criminisi algorithm,obtain better visual effect,and improve the quality of image inpainting.关键词
Criminisi算法/图像修复/可变大小样本块/模块相似性度量/颜色直方图Key words
Criminisi algorithm/image inpainting/variable-size exemplar block/module similarity measure/color histogram分类
信息技术与安全科学引用本文复制引用
王文豪,周静波,高尚兵,严云洋..Criminisi图像修复算法的优化[J].现代电子技术,2017,40(11):53-57,5.基金项目
国家自然科学基金(61402192) （61402192）
江苏省高校自然科学基金(14KJB520006) （14KJB520006）
江苏省淮安市科技支撑计划(HAG2013068) （HAG2013068）
