基于邻域一致性的数字X射线图像局部快速修复算法

摘要

This paper proposed an effective inpainting algorithm based on coherence transport for the industrial X-ray image low contrast and small defect areas which was similar to the background. Firstly, detecting the closed connected region, which was less than the threshold, was considered to be defective mask. Then,recorded the coordinates of the mask points to forming a closed connected region mask. Finally, used an non-iterative approach from outside to inside to complete the inpainting damaged image background. Experimental results show that the algorithm is better than FOE and PDE algorithms, and it can improve the quality of image restoration effectively and efficiency.%针对工业X射线扫描图像中存在微小异物及裂缝缺陷的特点,提出了一种基于邻域一致性的数字X射线图像局部快速修复算法.通过分析图像中封闭的连通区域的大小,设置蒙板并记录其坐标,采用从外到内、由粗到精的非迭代修复方法完成对扫描图像背景的修复过程.实验结果表明,与传统FOE和PDE修复方法相比,该算法在保证修复效率的前提下提高了图像的修复质量.