Multi-source Remote Sensing Image Fusion Related Technology Research

Multi-source remote sensing image fusion is one of important image processing technologies in multi-source remote sensing image information systems. With the rapid development of remote sensing technology, the application of multi-source remote sensing image fusion have been extended from military field to civil field, such as photogrammetry, resource reconnaissance, environment surveillance and so on. The fateful problem in the air of this field is how to implement high-powered image fusion. In this paper, we make research on the correlated technologies of multi-source remote sensing image fusion. Our research work focus on the following aspects:(1)An edge-based multi-source remote sensing image registration technique is proposed in this paper. We extract the edge feature of a reference image and a registering image using an algorithm of detecting modulus maxima in Contourlet subbands and the artificial edges and the inconsistency edges are removed based on an edge consistency check algorithm subsequently. The similarity measure of the edge feature between the reference image and the registering image is based on least trimmed square (LTS) Hausdorff distance and the space transform of image registration is estimated through the affine transform model simultaneously. Then, contraposing the existent prematurity problem of particle swarm optimization algorithm, a novel mutation operator based on student distribution is introduced to increase the population diversity and a modified adaptive particle swarm optimization algorithm is proposed. Finally, we implement high-powered multi-source remote sensing image registration through finding out the optimization space transform parameters using the modified adaptive particle swarm optimization algorithm.(2)An edge-oriented multi-source remote sensing image fusion method is proposed in this paper. On the basis of pixel-level multiresolution image fusion frame and region multiresolution image fusion frame proposed by G.Piella, we construct a feature-oriented multiresolution image fusion model firstly. Contraposing the problem of multi-source remote sensing image fusion, under the guidance of feature-oriented multiresolution image fusion model, a novel edge-oriented multi-source remote sensing image fusion method base on Contourlet transform is proposed. In the novel multi-source remote sensing image fusion method, the non-subsampled Contourlet transform is employed to decompose and to reconstruct the soure images and the edge feature information charactered by an image fuzzy edge field, which is constructed through Gauss distribution function, is inducted into the construction of subimages fusion rules in non-subsampled Contourlet transform field.(3)We make carefully research on the performance evaluation method of image fusion and then we make an all-around performance evaluation to the edge-oriented multi-source remote sensing image fusion method base on Contourlet transform proposed in this paper through selecting the appropriate objective performance evaluation indices. The experimental results show that the fused images obtained through the fusion method proposed in this paper possess favorable subject vision effect and objective evaluation performance. The fused images are propitious to object recognition and image understanding of the subsequent image processing step.