关键词: 数据融合, 均方差, 图像修复, 算法

Abstract: On the basis of analysis of several key image inpainting algorithms about implementation principle, applicability and superiority, considering their limited applicability, higher complexity and failured to take into account the structure information of damaged images, a new faster inpainting algorithm using structure and texture optimization was proposed, which is based on the fast marching method for weighted mean squared error and data fusion of Dempster-Shafer evidence theory applications. This technique can be used in reconstruction of damaged portions of ancient painting and also in removing entire objects from natural image. Through repairing the murals and natural images, the effectiveness of the proposed algorithm was verified by means of image completion and system simulation experiment. The proposed algorithm considers the structure information, achieves better repairing results and improves applicability.

Key words: data fusion, mean squared error, image inpainting, algorithm