The Research For Remote Sensing Image Fusion Based On Wavelet Transform

How to make use of high-resolution image and multi-spectral image fully, to improve definition and resolution of the multi-spectral image and retain the color feature of multi-spectral image, is the key point in the remote sensing field. The conventional approach called pixel-based fusion only based upon the pixel level, which take no account of spatial correlativity between the neighborhood pixels. The other method called region-based variance fusion neglected the feature of pixel structure. In addition, both of them considered only one feature as the final fusion judging basis, which is not very comprehensive. So they both lead to not good fusion results.In the background above, the dissertation is based on the remote-sensing image interpreter principal and wavelet MRA theory. Using the strongly matrix operation capability of Matlab software to extract the image information from TM and SPOT images ,and do the quantitative application. Integrated spatial and spectral information ,it proposed fusion methods called structrue similarity based on region and edge-enhanced fusion with wavelet's multi-resolution analysis. We apply these two algorithms to the typical cases of TM and SPOT, and compared the fusion results with the conventional approaches.