基于改进Hausdorff距离的图像配准方法

Image registration method based on improved Hausdorff distance

针对多尺度遥感图像灰度差异大的特点,利用特征集形状进行配准,提出了一种改进的Hausdorff距离及相应的图像匹配算法。首先采用基于尺度不变特征转换(scale-invariant feature transform,SIFT)的特征提取方法,提取多尺度图像间的尺度不变特征;然后利用Hausdorff距离作为适应度函数,通过遗传算法(genetic algorithm,GA)寻求图像间的几何变换参数;最后将待配准图像经过几何变换以及重采样与参考图像匹配,实现多尺度遥感图像的配准。实验结果表明,改进的Hausdorff距离算法与传统的Hausdorff相比,具有较高的配准精度和较快的配准速度,且稳定性和抗噪性更高,更适合用于图像配准。

In consideration of the features of remarkable difference in the gray-scale of the remote sensing image with multi-scales, this paper presents an image registration method with improved Hausdorff distance based on scale-invariant to solve the registration of multi -source remote sensing images. According to the method, the scale-invariant features of multi-scale images were firstly extracted by using the feature extraction method based on scale-invariant feature transform （ SIFT ） , and then the Hausdorff distance was used as the fitness function to seek for geometric image transformation parameters with the help of genetic algorithm（ GA） . At last,the image to be registered was re -sampled by using the transformation parameters and matched with the references image. The experimental results show that, compared with the traditional method of Hausdorff distance, the new method has higher registration accuracy and stability, and is more suitable for image registration.