Figure Pixel Level Fusion Method Of Multi-source Remote Sensing

At present, the image fusion technology in the military, medical science commerce, transport, environment and many other fields has been widely used. In this paper, several common traditional methods of image fusion and wavelet transform image fusion method have been studied. After a deep theoretical analysis of canonical correlation analysis (CCA) and multi-set canonical correlation analysis (MCCA),this paper proposed the basic strategies and methods of how to use CCA and MCCA for remote sensing image fusion.The main work of research and innovation are included as the following aspects:(1) Image registration. In this paper, we extracted keypoints with Forstner operator based on entropy theory and SIFT and then registrate image.For SIFT keypoints extraction algorithm, in order to accelerate the computing speed, this article built a pyramid with one octave and six layer instate of four octaves and five layers pyramid which was suggested by Lowe. After matching the keypoints, maybe it would come one-to-many points. The correlation coefficient is used as an evaluation factor to remove false matching points.(2) This papere achieved a number of common traditional image fusion methods, such as the fusion methods based on the weighted average, PCA, HIS, Brovey transform .Brovey method for improved integration was reasonable to enhance the three-band image brightness and clarity, reduce the degree of spectral distortion.(3) Deeply researching fusion methods based on wavelet transform. For the integration of different types of images we take different ways. It's in common that a reference image is introducd and high-frequency coefficients take the major principle factor; and in the selection of reference images and the low-frequency coefficients of the rational choice to adopt a different strategy.(4) In this paper, the solutions of how to fusio two images based on CCA-,and give the image fusion method of CCA with a combination of wavelet has been proposed. On this basis, the multiple canonical correlation analysis (MCCA) has been studied in-depth. And based on the principle of MCCA feature fusion, experiment has been done in the multi-spectral and multi-temporal remote sensing images. After that, this paper gave the analysis.