This paper aims to describe an improved method for integrating the SAR features into multispectral(MS) images based on the second generation curvelet transform. Curvelet transform(CT),as a special member of the emerging family of multiscale geometric transform, overcomes inherent limitation of traditional multiscale representation and owns very high directional sensitivity and anisotropy. It represents edges and singularities along curves much more efficiently than the traditional wavelet transform. To get more information in fusion image, the curvelet transform is introduced. The proposal of the second generation curvelet theory makes it understood and implemented more easily. SAR image has higher spatial resolution, but MS images have more spectral information. In order to get maximal integration of SAR features and the maximal preservation of the spectral content, two kinds of images may be implemented by different scales curvelet decomposition. And these decomposed curvelet coefficients can be processed according to certain fusion regular. Then detail information of SAR and approximate information of MS will be extracted by inverse curvelet transform respectively. At last fused image is obtained by injecting SAR's detail information into the MS's approximate information. Landsat TM and SAR images covering a region of sanshui in Guangdong province are used to evaluate the effect of the proposed method and some other fusion methods in terms of spectral preservation and spatial resolution improvement. The results show that the proposed method can provide richer information in the spatial and spectral domain.
The research presented in this paper is aimed at the development of multisensor image fusion. The proposed approach is suitable for integration pan-sharpening of multispectral (MS) bands and SAR imagery based on intensity modulation through the a-trous wavelet transform (ATWT) and the curvelet transform(CT). The ATWT is suitable for dealing with objects where the interesting phenomena, e.g., singularities, are associated with exceptional points, and CT as a new multiscale geometric analysis algorithm is more appropriate for the analysis of the image edges and has better approximation precision and sparsity description. This proposed fusion algorithm makes full use of advantages of these multiscale analysis tools, thus it extracts SPOT-Pan high-pass details from the panchrmomatic image by means of the ATWT and SAR texture and edges by details and rationing the despeckled SAR image to its lowpass approximation derived from the CT.SPOT-Pan high-pass details and SAR texture and edges are used to modulate intensity derived from IHS transform of MS bands. SPOT-Pan, Landsat-MS and Radarsat-SAR images covering a region of sanshui in Guangdong province are used to evaluate the effect of the proposed method. The experiment result shows that the proposed algorithm has greatly improved spatial resolution while it keeps the spectral fidelity.