基于多模板方向的图像修复方法

在基于样本图像修复的基础上,针对采用单一方向模板的Criminisi算法,在查找最佳匹配块过程中产生的错误匹配甚至查找失败问题,提出了一种基于自适应模板方向的图像修复方法。该算法根据自然图像中存在大量对称或有规则物体的特点,在进行模板的搜索时,通过对待修复块或样本块进行不同角度的旋转和/或映射变换,增大样本空间,来提高最佳匹配块的搜索成功率,防止错误匹配的发生。仿真实验结果证明,该方法能有效地改善图像的修复效果。

Image Inpainting Algorithm Based on Adaptive Template Direction

On the basis of exemplar-based inpaiting, an adaptive template direction algorithm was proposed to solve the problem that Criminisi algorithm easily leads to mismatching problems in the process of searching the best match patch. The selection of the direction of the template matching may raise the success rate of finding the best match patch and prevent mismatches. The simulation results show that this algorithm is effective in improving the image inpainting effect.