中文摘要：

传统基于样本块的图像修复方法只能在平移空间中进行搜索,当出现旋转和尺度变化时无法实现匹配块的最优选取.为解决上述问题,将Patchmatch 快速搜索算法引入图像修复领域;在此基础上,通过改进能量函数来实现旋转及尺度空间拓展;最后利用LM 优化算法获得最优化模型参数.仿真实验结果表明,当破损图像存在旋转及尺度变换时,该方法仍然能够准确找到其最优匹配块,取得理想的修复效果.

英文摘要：

In the traditional examplar-based image completion algorithm,searching is limited in translation space, thus it is hard to choose the optimal patch when rotation and scale transformation exist in image. In order to solve this problem,in this paper,Patchmatch rapid search method is introduced intothe image completion field. On this base,the rotation and scale space expansion is realized with the improved energy function. Finally the LM optimiza-tion method is adopted to obtain the optimal transformation parameters. The simulation results show that the method can obtain the optimal patch even when rotation and scaling transformation exist,and achieve perfect completion results.