Fusion of Multi-spectral and Panchromatic Images Using Optimal Estimation Theory

Considering panchromatic image and multi-spectral images contain different spectral information and same spatial information with different observation accuracy,thus the optimal spatial component information could be estimated.A new image fusion method is proposed based on least-squares estimation theory and wavelet transformation.After the remote sensing images are decomposed by wavelet transform,the method uses the prior information of remote sensing image: the ratio of spatial resolution to determine the weights in the images fusion.Image fusion is completed by the least squares theory and wavelet reconstruction finally.Compared with the previous methods that spatial information of multi-spectral images is replaced completely or additive model,it is more accurate and robust.Two data sets are used to evaluate the proposed fusion method,QuickBird multi-spectral images and panchromatic image,TM multi-spectral images and SPOT panchromatic image.The fusion results are evaluated visually and statistically.The evaluation shows that relative to the other fusion method based on wavelet transform,the proposed method achieves the best fusion images in obtaining high spatial resolution of panchromatic image and preserving spectral properties of multi-spectral images.