Abstract: In order to improve the problem of the lack of detailed information in expression of image
using non-subsampled Contourlet transform (NSCT), this paper proposes an improved method based
on principal component analysis (PCA) and NSCT transform remote sensing image fusion. Firstly,
PCA transform is applied to the low spatial resolution multi-spectral (MS) image, and then the first
principal component (PC1) is extracted. Secondly, NSCT transform is applied to the PC1 and the high
spatial resolution panchromatic (PAN) image. For the low frequency coefficients of the above two,
the rules of wavelet transform fusion are used, and for the high frequency coefficients the adaptive
weighted fusion rules based on region standard deviation are used. Finally, we get the fusion image by
using inverse NSCT transform and inverse PCA transform. The results show that the method
combines the detail information of the source image effectively, and also get better visual effect and
better evaluation index.

Key words: remote sensing image fusion, NSCT transform, PCA transform, wavelet transform, fusion
rules, region standard deviation of the adaptive weighted