Automatic Multi-spectral Image Registration for Tiangong-2 Wide-Band Imaging Spectrometer

Abstract

Automatic multi-spectral image registration can improve the application efficiency and ability of remote sensing images while this technology is yet to be improved. Based on the remote sensing data of Tiangong-2 wide-band imaging spectrometer, the differences and similarities between multi-spectral images were analyzed in this paper. An edge extraction method with edge symmetry and non-directionality was proposed according to the neighborhood structure of image pixels. Using the edge information images, multi-spectral image registration was achieved through the measurement of mutual information. A contrast experiment was also set up to compare the effects of the proposed edge extraction method with those of several traditional edge extraction methods in image registration. The results showed that the registration precision of this method is higher, indicating that the extracted edge features were more stable among multi-spectral images.