基于图像配准的潜在痕迹提取

Recovering latent marks in distracting background based on registration techniques

介绍了依据图像配准技术在复杂背景中提取潜在痕迹的一种方法。图像配准是图像处理的一个重要的基础环节,传统的图像配准算法存在着各种不确定性及计算复杂性,而且特征点的选择困难且自动性差。而我们要实现从复杂的背景中提取潜在的痕迹必须要求图像要精确对准到亚像素级。针对传统算法中存在的问题,分别采用了一种从特征点提取、匹配到图像间点变换估计及光学变换估计的鲁棒算法。这些算法经过实验证实是鲁棒的且是实用的。

In this paper, a method which uses the registration techniques to allow latent mark to be recovered from the images with a distracting background patterns is introduced. Image registration is an essential step in image processing. A robust algorithm of image registration ,which can overcome the restrictions of existing methods and gains sub - pixel precision for image registration are presented. The experiments prove the method is stable very much. It is high valuable in practice,