Abstract: [Purposes]To address the challenge of simultaneously achieving high spatial detail and rich spectral information for traditional remote sensing image fusion methods,this paper proposes a fusion method based on high-frequency detail enhancement and low-frequency feature optimization.[Methods]Firstly,the 3D and 2D Curvelet transforms are used to decompose multi-spectral and panchromatic images to obtain the corresponding high and low frequency coefficients.Then,a component substitution strategy is employed to construct the fusion rule for high-frequency coefficients,and a band-adaptive adjustment coefficient is introduced to optimize the high-frequency fused coefficients.For the low-frequency fusion,a rule is established based on the similarity features between the multi-spectral and panchromatic low-frequency coefficients.Additionally,the Laplacian features of the panchromatic low-frequency coefficients are used to enhance the contour information of the multi-spectral image,further optimizing the low-frequency fused coefficients.Finally,the optimized high-frequency and low-frequency fused coefficients are reconstructed through the inverse 3D Curvelet transformation to obtain the fused multi-spectral image.[Findings]Experimental results on GF-2,QuickBird and WorldView-3 remote sensing images demonstrate that the proposed method achieves significant improvements in both visual perception and quantitative metrics.[Conclusions]It effectively enhances image sharpness and detail representation while maintaining high spectral fidelity.

Key words: image fusion, 3D Curvelet transformation, Laplacian features, similar features, spectral fidelity