摘要
描述了一种基于H ausdorff度量的合成孔径雷达和光学图像配准方法。首先用基于低帽滤波的方法提取待配准图像的闭合轮廓。然后对较长的轮廓进行H ausdorff度量初匹配,并对初匹配的结果使用轮廓中心的相对距离比直方图聚束检测法进行一致性检测。最后,在得到正确的闭合轮廓对后,使用最小二乘法计算图像的变换参数。考虑到雷达图像的相干斑噪声以及多传感器图像成像时间造成的变形,多传感器图像提取的轮廓会有一定的差别。而H ausdorff度量对误差有很好的容忍性,因此本方法可以对多传感器图像进行配准。
This paper presents a multi-sensor image registration approach based on Hausdorff distance. Firstly, close contours of images are extracted through bottom-hat filter and threshold. Secondly Hausdorff distances between the longer contours of different maps are calculated, and contour pairs with small enough Hausdorff distance are deemed matching contours pairs. Then the consistency checking succeeds by relative distance ratio histogram clustering based on center of contours. Finally accurate contour pairs are obtained so that image transformation parameters are attained by applying LSM in interrelated parts. Considering speckle noise of SAR images and distortion due to multi-sensor images being taken at different time, extracted contours of SAR and optical images differ to certain extent. As Hausdorff distance allows large contour difference, the approach presented in this paper can register images of wide translational and distortional range.
出处
《遥感技术与应用》
CSCD
2006年第5期473-476,共4页
Remote Sensing Technology and Application
