结合多尺度角点和Harris变换的遥感图像配准方法

A Remote Sensing Image Registration Algorithm Based on Multi-scale Corner and Harris Transformation

提出了一种新的基于多尺度角点和Harris变换相结合的角点检测的图像配准方法。其核心思想是采用Harris算子和Laplace图像金字塔相结合,对原始图像进行不同尺度的向下采样并对得到的图像进行角点检测,得到角点的角度和尺度响应的最值点比(尺度精定位),对图像进行旋转和平移变换,完成图像配准。理论分析和仿真实验表明,该算法具有较高的配准精度且计算量较小。

A new image registration algorithm based on multi-scale corner and Harris transformation is proposed,whose core idea is the combination of Harris operator and the Laplace image pyramid.The original image is sampled down on different scales and the corner of the obtained image is detected.The angle of the corner and the ratio of maximum values of the scale response(scale fine positioning) are obtained.Then the image is rotated and translated to complete the image registration.Theoretical analysis and simulation results show that the algorithm has higher registration accuracy and less computation amount.