摘要
针对图像缺损修复需求设计了一种修复算法.该算法以Criminisi算法为基础,引入标记符控制分水岭算法对图像按景物进行分块,提高图像缺损部分修复的准确率;根据最近最佳匹配原则修复图像,引入像素距离远近控制函数以减少搜索匹配块用时;引入图像局部色彩方差,优化Criminisi算法的优先权计算方式;引入协方差,提高填充块与待修复区域图像的匹配度;重新设置图像修复的置信度公式,减小因置信度不断更新时放大的误差.结合多种类型的缺损图像实例进行修复验证,均取得了很好的修复效果.实验结果表明:与其他算法对比,该算法大幅度减少了缺损图像修复用时,更有效地恢复了图像原有信息.
In this study,we designed a repair algorithm to satisfy the requirement for repairing the defects in images.Based on the Criminisi algorithm,we introduced the marker-controlled watershed algorithm to segment an image into parts based on the scene for improving the image defect repair accuracy.Further,based on the nearest best matching principle,the image was repaired and a control function was introduced with respect to the pixel distance for reducing time for searching matching blocks.Then,we optimized the priority calculation method of the Criminisi algorithm by introducing the local color variance,and improved matching degree between the filled segments and the sections to be repaired based on the covariance.Subsequently,the confidence formula with respect to image repair was reset to reduce the magnifying error when the confidence value was continuously updated.Additionally,we considered many defect images as examples to verify the repairing effect,and the obtained results demonstrated that the designed algorithm has an excellent repair effect.Experimental results show that compared with other algorithms,the proposed algorithm can considerably reduce the time required to repair the damaged image and more effectively restore the original image information.
作者
王永飞
Wang Yongfei(School of Computer Science and Technology,Anhui University,Hefei,Anhui 230601,China;Department of Information Technology,Tongling Polytechnic,Tongling,Anhui 244061,China)
出处
《激光与光电子学进展》
CSCD
北大核心
2020年第12期106-116,共11页
Laser & Optoelectronics Progress
基金
安徽省高等学校质量工程项目(2017mooc348)。
关键词
图像处理
图像缺损
图像景物分块
划痕图像修复
图像景物去除
image processing
image defects
image scenic segmentation
scratch image repair
image scene removal
