Pansharpening of multispectral images using the nonseparable framelet lifting transform with high vanishing moments

摘要

This paper proposes a novel nonseparable lifting scheme for wavelet frames with high vanishing moments. A specific nonseparable framelet lifting transform (NFLT), combined with a modified covariance intersection (CI) algorithm, has been applied to pansharpening of multispectral images. Experiments are carried out on the multispectral and panchromatic images acquired by the SPOT, QuickBird and Landsat spaceborne sensors. Benefiting from the high order of vanishing moments, the proposed NFLT can distinguish the low- and high-frequency efficiently and can compact most of the energy into the low-pass subband. Thus the spectral distortion can be minimized. Experimental results show that the NFLT-CI method reduces the spectral distortion while improves the spatial resolution simultaneously, and outperforms the other state-of-the-art methods derived from various transforms and injection models.