Pan-sharpening aims at integrating spectral information from a multi-spectral (MS) image and spatial information from a panchromatic (PAN) image in a fused image with both high spectral and spatial resolutions. Numerous pan-sharpening methods are based on intensity-hue-saturation (IHS) transform, which may cause evident spectral distortion. To address this problem, an IHS-based pan-sharpening method using ripplet transform and compressed sensing is proposed. Firstly, the IHS transform is applied to the MS image to separate intensity components. Secondly, discrete ripplet transform (DRT) is implemented on the intensity component and the PAN image to obtain multi-scale sub-images. High-frequency sub-images are fused by a local variance algorithm and, for low-frequency sub-images, compressed sensing is introduced for the reconstruction of the intensity component so as to integrate the local information from both the intensity component and the PAN image. The specific fusion rule is defined by local difference. Finally, the inverse ripplet transform and inverse IHS transform are coupled to generate the pan-sharpened image. The proposed method is compared with five state-of-the-art pan-sharpening methods and also the Gram-Schmidt (GS) method through visual and quantitative analysis of WorldView-2, Pleiades and Triplesat datasets. The experimental results reveal that the proposed method achieves relatively higher spatial resolution and more desirable spectral fidelity.
Keywords: image fusion; intensity-hue-saturation transform; remote sensing; ripplet transform; sparse representation.