Pan-sharpening of QuickBird satellite images using multiresolution techniques: wavelets, contourlets, and curvelets

Abstract

The Pan-sharpening technique is of special interest in remote sensing because satellites usually take panchromatic (PAN) and multispectral (MS) images separately; therefore, later fusion is necessary. In the present study, we focus on the satellite QuickBird because its images are among those that have the highest spatial resolution and are useful for regional analysis and high-resolution thematic cartography. Also, we perform the fusion task by multiresolution techniques, as numerous studies have demonstrated that the wavelet method (the origin of multiresolution methods) is one of the best for Pan-sharpening. Nevertheless, other multiresolution methods which inherited the wavelet concept, such as contourlet and curvelet, have scarcely been used in Pan-sharpening. In our study, we use undecimated algorithms in wavelets and contourlets while using a decimated algorithm for curvelets. The decimated algorithm presents an inhomogeneous colour distribution at a particular resolution level. We found that, for each method, more than one resolution level is suitable for Pan-sharpening. On the other hand, all methods at all the resolution levels present good spatial resolution when they are evaluated visually. Nevertheless, there are spectral differences between methods and resolution levels. Finally, from the analysis and discussion of the results, we recommend two methods for their functionality, these methods being the wavelet at resolution level 2 and the contourlet at resolution level 4.