Contourlet-Based Multispectral Image Fusion Using Free Search Differential Evolution

In this paper, the multispectral image fusion task is converted into an optimisation problem, to satisfy the objective of maximal injection of spatial information with minimal spectral distortion. Contourlet transform (CT) is employed to extract the spatial high-frequency coefficients from PAN image and they are weighted and injected into each band of the corresponding components of multispectral data. The weighted coefficients are found by using the advanced evolutionary intelligence technique called free search differential evolution (FSDE). The novelty of this paper is to introduce FSDE for improved application of CT for image fusion. The proposed method, CT-FSDE, was tested and compared with principal component analysis (PCA), Laplacian pyramid (LP), wavelet transform (WT), and CT over a WorldView-2 dataset. In order to study the effectiveness of FSDE, I also compared it with two advanced evolutionary algorithms, JADE and PS2O, which were developed from differential evolution and particle swarm optimisation, respectively. The quantitative results from conducted experiments show that the proposed method provides high-quality spatial details and also preserves spectral information well, which highlights the benefits of the proposed method for multispectral image fusion.