目标移除的Criminisi图像修复算法

Criminisi Image Restoration Algorithm for Object Removal

待修复区域标记、优先权、最佳匹配模块的搜索及填充和更新置信度是Criminisi图像修复算法的4个重要组成部分,每一部分对Criminisi算法的图像修复结果都有不可忽视的作用。针对大目标移除的Criminisi图像修复,从数学的角度优化待修复区域标记和提高优先权可信度,并将蝙蝠算法运用到最佳匹配模板搜索中,提高修复效率。实验表明:本文改进的算法能够有效的减少错误信息的累积,具有较好的实用性。

Criminisi algorithm consists of four steps which are marking area, the priority, the best sample patch and updating the degree of confidence and each step has the important effect. For removing a big target, this paper uses the reasonable marked and improved priority, and then the bat algorithm is applied to search the best match template in Criminisi image restoration. Experiment results show that this method can reduce the inaccurate matching and has relatively high practical value.