FRAMELET-BASED MULTIRESOLUTION IMAGE FUSION WITH AN IMPROVED
INTENSITY-HUE-SATURATION TRANSFORM
M. J. Choi *, D. H. Lee, H. S. Lim
Satellite Information Research Institute, KARI, 45, Eoeun-dong, Yuseong-gu, Daejeon, 305-333, KOREA
prime, dhlee - hslim@kari.re.kr
WG VII/6 - Remote Sensing Data Fusion
KEY WORDS: Satellite remote sensing, Pan-sharpening, Intensity-Hue-Saturation transform, Wavelets, Framelets, IKONOS
imagery.
ABSTRACT:
The fusion of the panchromatic (PAN) image with multispectral (MS) images is a very useful tool for remote sensing applications
that require both high-spatial and high-spectral resolution. A hybrid fusion method, which comprises of the standard intensity-hue-
saturation (IHS) transform and a wavelet transform, was recently introduced in the literature. The hybrid fusion method was able to
eliminate drawbacks of the IHS and wavelet-based methods, while keeping the advantages. In this paper, we introduce a new hybrid
fusion method based on the fast IHS transform with a control parameter along with the framelet transform. The control parameter is
used to minimize the radiance difference between the PAN image and the intensity image. In addition, the framelet transform, which
is a natural generalization of orthonormal wavelets if redundancy is introduced into the wavelet system, is used to extract the
detailed spatial information from the difference image of the PAN image and the intensity image. In order to verify the performance
of the proposed method, an IKONOS PAN image with MS images were used. The proposed method produces more satisfactory
results, both visually and quantitatively.
1. INTRODUCTION
Due to the technical limitation, most earth observe satellites,
such as the SPOT, IKONOS, QuickBird, and KOMPSAT-2,
provide a bundle of multispectral (MS) images at a lower
spatial resolution and a panchromatic (PAN) image at a higher
spatial resolution, instead of MS images at a high spatial
resolution. Therefore, a pan-sharpening technique is very useful
tool for remote sensing applications that require both high-
spatial and high-spectral resolution (Zhang,2004).
Many image fusion techniques and software tools have been
developed for specific applications. Of the hundreds of varied
image fusion techniques, the best known methods are the
intensity-hue-saturation (IHS) transform, principal component
analysis, arithmetic combinations, and wavelet-based fusion
methods. However, most of the existing techniques which
perform suitably well with medium-resolution images, cannot
always satisfy the fusion of MS and PAN high resolution
images.
Recently, Gonzalez-Audicana et al. (2004) introduced a hybrid
fusion method comprised of the standard IHS transform and an
undecimated discrete wavelet transform (UDWT). The hybrid
fusion method was able to eliminate drawbacks of the IHS and
wavelet-based methods, while keeping the advantages.
Nevertheless, it is not efficient enough to quickly merge
massive volumes of data from high-resolution satellite images
because of its high computational complexity.
To reduce the computational cost, a fast IHS (FIHS) transform
introduced by Tu et al. (2004) is employed on a new hybrid
fusion method. Aside from its fast computing capability for
fusing images, this method can extend traditional three-order
transformations to an arbitrary order. In addition, we employed
a framelet transform which is a natural generalization of
orthonormal wavelets if redundancy is introduced into the
wavelet system. It thereby avoids some of artifacts that arise
when the decimated discrete wavelet transform is applied to the
image fusion like the UDWT. Moreover, it can reduce the
computational cost to a half of that of the UDWT.
More importantly, the proposed hybrid method enables us to
reduce the overall computational cost to a half one again
through simplifying the original mathematical model. A
detailed explanation will be given later in the body of this paper.
On the other hand, one of disadvantages of IHS-based methods
is that they destroy the spectral characteristics of the MS data
(Choi,2006). Recent studies have shown that the large
difference between the values of the PAN and intensity images
appears to cause the large spectral distortion of fused MS
images. Indeed, this difference (PAN-I) causes the altered
saturation component in the RGB-IHS conversion model.
Therefore, a FIHS transform with a control parameter is
employed to minimize the radiance difference between the PAN
image and the intensity image.
In order to verify the performance of the proposed method, an
IKONOS PAN image with MS images were used. Then the
spatial and spectral quality of the resulting images were
analyzed and compared to images obtained from other well-
known methods. The following chapter will explain the theory
of the framelet transform. Readers unfamiliar with the wavelet
theory are advised to continue to the third chapter, which the
Corresponding author.