Abstract: There exits the structure discontinuous phenomenon when the strong structure texture image is inpainted by exemplar-based image inpainting algorithm. In view of this problem, this paper presents an improved algorithm for exemplar-based image inpainting. It introduces structure tensor in computing data term, and realizes anisotropic linear structure to determine the inpainting order of the target patch. The priority is the sum of confidence term and data term. Experimental results show the improved algorithm is effective in inpainting structured texture images.

Key words: image inpainting, texture synthesis, exemplar, structure tensor

摘要： 基于样例的图像修复算法在修复强结构纹理图像时存在结构不连续现象。针对该问题，提出一种基于样例的图像修复改进算法。在计算数据项时引入结构张量，实现各向异性线性结构的优先级，以决定目标区域修复的先后顺序，使用置信度项和数据项加权和的方式计算优先级。实验结果证明，改进算法对强结构纹理图像的修复效果较优。

关键词: 图像修复, 纹理合成, 样例, 结构张量

CLC Number: 

• TP391