基于轮廓的多源图像的配准

A Research of Contour - based Registration Technique for Multi - Source Images

不同源图像间由于灰度和分辨率水平上的显著差异,使得两图像间的配准问题十分困难。为此,本文避开了常规所采用的直接灰度相关找配准点,首先提取各图像中的整体轮廓,再对轮廓作链码相关,进而找出配准点。实验结果表明,基于轮廓的配准技术应用在不同源图像间配准时明显要优于直接灰度相关法,而且由于链码的应用,计算量也大为减少,在图像融合匹配技术领域具有重要意义。

Because of difference on the gray - level and resolution - level between multi - source images, it is very difficult to register two images. To avoid the direct correlation on the two images, a contour - based registration approach to the multi - source images is presented in this paper. We first extract the overall contours in the images and encode them into the chain code, then compute the cross - correlation based on their chain code to get the correct registration point. The validity and calcultional efficiency of this method are shown by the experimental results.