A Low Computational-Cost Method to Fuse IKONOS Images Using the Spectral Response Function of Its Sensors

Abstract

Probably the most popular image fusion method is that based on the intensity-hue-saturation (IHS) transform. Although the spatial enhancement of the IHS-merged images is high, the distortion of its spectral information may also be important. In recent years, several methods have been developed to minimize this problem, being those based on wavelets widely used. However, the high computational cost of these approaches makes them unattractive to applications that involve fast merging of very large volumes of data. In this paper, we present a low computational-cost image fusion method based on the fast IHS transform, which uses the information of the spectral response functions of the low-resolution multispectral (LRM) and high-resolution panchromatic (HRP) sensors to minimize the spectral distortion problem. Using this information, we directly obtain from the HRP image the intensity image that the LRM sensor would observe if it worked at a spatial resolution similar to that of the HRP image. The experimental results carried out on IKONOS images demonstrate that the proposed approach can perform as well as wavelet-based approaches with a lower computational cost.