Fusion of Multispectral and Panchromatic Images Using a Restoration-Based Method

• Published in: IEEE transactions on geoscience and remote sensing 2009-05, Vol.47 (5), p.1482-1491
• Main Authors: Zhenhua Li, Leung, H.
• Format: Article
• Language: English
• Subjects: Algorithms; Applied geophysics; Bands; Closed-form solution; Earth sciences; Earth, ocean, space; Exact sciences and technology; Exact solutions; Humans; Image fusion; Image restoration; Internal geophysics; Mathematical analysis; Mathematical models; multispectral image; Multispectral imaging; Object detection; panchromatic image; Remote sensing; Satellites; Sensor fusion; Spatial resolution; Wavelengths
• ISSN: 0196-2892; 1558-0644
• DOI: 10.1109/TGRS.2008.2005639

Many remote-sensing satellites can obtain images in multispectral and panchromatic bands. By fusing low-resolution multispectral and high-resolution panchromatic images, one can obtain high-resolution multispectral images. In this paper, an image fusion algorithm based on image restoration is proposed to combine multispectral and panchromatic images. For remote-sensing satellites, the wavelength of the panchromatic band usually covers the wavelengths of the multispectral bands. This relationship between the two kinds of images is useful for fusion. In our approach, the low-resolution multispectral images are first resampled to the scale of the high-resolution panchromatic image. The relationship between these two kinds of images is then used to restore the resampled multispectral images. That is, the resampled multispectral images are modeled as the noisy blurred versions of the ideal multispectral images, and the high-resolution panchromatic image is modeled as a linear combination of the ideal multispectral images plus the observation noise. The ideal high-resolution multispectral images are then estimated based on the panchromatic and the resampled multispectral images. A closed-form solution of the fused images is derived here. Experiments show that the proposed fusion algorithm works effectively in integrating multispectral and panchromatic images.