Optical/SAR image registration based on cross-correlation with multi-scale and multi-direction Gabor characteristic matrixes

Optical/SAR (Synthetic Aperture Radar) image registration is one of the key technologies in image fusion or image guidance with SAR imaging. Generally we implement characteristic extraction and scene matching on image registration with optical and SAR images. This paper presents a new heterogeneous optical/SAR image registration method. Firstly we implement directional Frost filtering on SAR image so that we can reduce coherent speckle and keep the details of the edges and textures. Secondly we carry out the Gaussian gradient image of the optical and SAR image respectively. Then we extract characteristics from the two Gaussian gradient images using two-dimensional multi-scale and multi-direction Gabor filters respectively and then form two Gabor characteristic matrixes. Finally we implement NCC (normalized cross-correlation) matching on the two characteristic matrixes. The experiment result shows that the heterogeneous image registration method mentioned in this paper has preferable robustness and accuracy.