Remote Sensing Image Fusion Methods Based On Multi-scale Geometric Transform And Particle Swarm Optimization

With the development of earth observation technology by remote sensing satellites,more and more remote sensing image data are provided by different kinds of remote sensors.So the remote sensing image fusion technology appears.This technology can integrate the multi-source remote sensing data with different features into a new image with new features according to certain fusion rules.In the fusion process,the complementary information between multiple images is preserved,and the redundant features are eliminated,which realizes the complementary advantages of multi-sensor data.The fusion technology for remote sensing image extends the application range of remote sensing image data,improves the practical value of remote sensing image data,and provides a reliable basis for the follow-up analysis,interpretation and understanding of remote sensing images.We proposed two remote sensing image fusion algorithms based on the ideal of further investigation on the multi-scale geometric transformation theory,the particle swarm optimization algorithm(PSO)and the spatial domain fusion.Two following works have been completed:(1)The particle swarm optimization algorithm(PSO)was introduced into the Contourlet transform coefficients by analyzing the characteristics of the Contourlet transform coefficients and the optimization efficiency of the particle swarm optimization algorithm.Firstly,the luminance component of panchromatic image and multispectral image was transformed by Contourlet transform on the basis of IHS transform.Secondly,according to the different characteristic information between the decomposed low frequency and high frequency coefficients,we regarded the difference between information entropy and relative deviation as fitness function and used particle swarm optimization algorithm(PSO)to adaptively find the optimal weighted coefficient for fusion.Then,the structural similarity was used as the objective function of particle swarm optimization algorithm(PSO)on the high frequency coefficient to find the optimal threshold of the similarity among the high frequency sub-bands,and images were fused on the basis of the fusion rules for structural similarity.Finally,the fused images were obtained by inverse transform.The simulation experiment results show that the proposed algorithm can effectively preserve the spectral information and spatial information of the original images,which is an effective and feasible remote sensing image fusion method.(2)According to the distribution of edge features and non edge features on remote sensing images,combining the effectiveness of Canny operator towards edge extraction,we presented a remote sensing image fusion algorithm by coupling edge detection and optimization method based on non-subsampled Contourlet transform(NSCT).Firstly,the edge information of the panchromatic image was distinguished to the non-edge by Canny operator on the basis of IHS transform,and the edge of the panchromatic image was enhanced by fusing the panchromatic image and I component of multi-spectral image according to the characteristics of edge distribution.Then,the enhanced panchromatic image and I component of multi-spectral image were respectively decomposed by NSCT,the method of selective weighted summation was used in the lowpass sub-band,the PSO was performed to search the best threshold p and the fusion rules of regional structure similarity was used in the highpass sub-band.Finally,the fused image was reconstructed by inverse transform of non-subsampled Contourlet transform(NSCT)and IHS.The simulation experiment results show that the distribution of edge features in fused image are obvious,it has more spectral signature information and obtains a better visual effect.In addition,the quantitative index in the fused image has been enhanced.