面向图像修复的域相似算法

Similar neighborhood approach for image inpainting

基于偏微分方程(PDE)的图像修复和基于纹理合成的图像修复是目前数字图像修复中的重要方法,虽然均能较好地修复图像,但是修复的效率较低。提出了一种采用域相似修复图像的新算法,先对待修复区域边界上的所有待修复点计算优先级,然后按照优先级从大到小的顺序修复图像;该算法以像素点邻域的相似来衡量两个像素点相似的程度,充分考虑了待修复像素的邻域中已知信息对该像素的影响。仿真实验结果表明,该算法不仅能较好地修复图像,而且在同等修复区域和修复效果的条件下具有更高的修复效率。

Image inpainting based on partial differential equation and image completion based on texture synthesis is a very important method in digital image inpainting, in the present, these methods perform well, but they consume much time. This paper presents a new image inpainting algorithm based on the similarity of neighborhood. Firstly, the paper calculates the priority of every boundary point and makes the descending priority order as an inpainting order, the algorithm takes the similarity of pixel’s neighborhood to measure the degree of the similarity of pixel, it fully considers the influence of known information of neighborhood for pixel, then, calculates the average of the known pixels whose meet the requirements of similarity as the values to be filled point. Experimental results show that this algorithm can inpaint image better, and under the same scope of to be filled and effect of inpainting conditions, this algorithm has higher inpainting efficiency.