基于K-L变换的两传感器图像自动配准

基于合成孔径雷达(SAR)图像和光学图像间的变形可用仿射变换来近似,提出了一种SAR图像和光学图像自动配准的方案.对分割后的图像进行K-L变换得到两幅图像的主分量与坐标轴之间的关系,从而得出两幅图像的角度关系,接着,利用旋转后图像的横纵象素范围的比值和前景重心偏移值得到仿射变换的缩放和平移参数,最后再利用二值图像相关对仿射变换的结果图像进行精配,从而实现较高精度的图像配准.实验结果表明,在满足精度要求的情况下,本方法可以自动完成具有明显方向性和较大前景目标图像的配准任务.

Automatic bi-sensor image registration based on K-L transformation

Since the distortion between synthetic aperture radar (SAR) and optical images is almost an affine transform. In this paper, K-L transformation is applied on segmented images to determine the relationship between the principal components of both images and the coordinate, thus the rotational angle between images is acquired. Then the ratio of extent in vertical and horizontal directions and the centroids position of foreground area are used to determine the scaling and translational parameters respectively. Finally, the affine-transformed result is refined by binary image correlation to achieve high precise image registration. Experimental results indicated that this method can perform automatic registration under precision acquirement especially for images with orientation and large foreground area.