数学形态学在Criminisi图像修复算法中的应用.pdfVIP

第37 卷 第7 期 红 外 技 术 Vol.37 No.7 2015 年7 月 Infrared Technology July 2015 数学形态学在Criminisi 图像修复算法中的应用 李 尊，吴 谨，刘 劲 （武汉科技大学信息科学与工程学院，湖北 武汉 430081 ） 摘要：Criminisi 图像修复算法的修复步骤由待修复区域的标记、优先权的计算、最佳匹配块搜索与填 充和更新置信度4 部分组成。其中待修复区域的标记是进行Criminisi 图像修复算法的基础，因此合理 的标记待修复区域是提高Criminisi 图像修复算法修复效果的有效路径。引入数学形态学对待修复图像 进行处理，即利用腐蚀与膨胀组合，对待修复区域边缘进行处理，进而进行Criminisi 图像修复算法操 作。实验结果表明经数学形态学处理后的标记可以降低错误信息的累积，大大提升了图像修复的质量， 具有较高的实用价值。 关键词：数学形态学；Criminisi 算法；图像预处理 中图分类号：TP391.41 文献标识码：A 文章编号：1001-8891(2015)07-0574-05 The Application of Mathematical Morphology in the Criminisi Algorithm of Image Inpainting LI Zun ，WU Jin ，LIU Jin (College of Information Science and Technology, WuHan University of Science and Techology, Wuhan 438001, China) Abstract ：Criminisi algorithm consists of four steps which are marked area. The priority, the best sample patch and updating the degree of confidence, marked area is the basis of Criminisi algorithm, so it is a effective way to improve the effect of Criminisi algorithm. In order to obtain the reasonable marked area, this paper used mathematical morphology in edge, which is corrosion and inflation, and then used Criminisi algorithm of image inpainting. The experiment shows that the marked by mathematical morphology can reduce the accumulation of error messages and improves the quality of the image restoration greatly which has high practical value. Key words ：mathematical morphology ，Criminisi algorithm，image pre-processing 轻图像修复效果人为痕迹的问题；文献[6]引入曲率到 0 引言