234
A NEW PAN-SHARPENING ALGORITHM AND ITS APPLICATION
IN GEOGRAPHIC FEATURES INFORMATION EXTRACTION
ZHU Lijiang 3
a Beijing Research Institute of Uranium Geology, RO.Box 9818,Beijing 100029,China
E-mail: zljrenwen@hotmail.com
KEY WORDS: Pan-sharpening, PSF, Fusion, Quickbird
ABSTRACT:
Many different fusion techniques were developed for fusing multi-spectral low-resolution remotely sensed images with a more highly
resolved panchromatic image, the goal is to obtain a high-resolution multi-spectral image which combines the spectral characteristic
of the low-resolution data with spatial resolution of the panchromatic image. Almost all these fusion methods are not satisfactory in
tiny geographic features (road, building, etc) information extraction for some very highly resolution satellite imagery collected from
the sensor such as Quickbird, Ikonos and so on. In this paper, we analyzed the process of image data collection and found that two
factors caused by the sensor’s orientation to the scene is the main factor that cause a misregistration between the pan and MS bands
and result in a blurry that is apparent around bright objects in fusion process. One is parallax (different look angles) between the
bands causes misregistration before the fusion. Another is the diffraction of the different MS bands cannot be neglected because of
the image have so high resolution and different wavelength among the bands. Through constructing a PSF model between different
pixel and bands in same satellite scene, the new fusion method solves these problems according to the geometry relation between the
scene and the sensor. As an example, the new pan-sharpening algorithm and other data fusion methods like HIS, PCA and UNB
(Digital Globe’s default pan sharpening algorithm) were applied to four related multi-spectral images of a Quickbird satellite scene to
extract geographic features information. The comparison between these pan-sharpening algorithms shows that the new
pan-sharpening algorithm is perfect in the test scene.
1. INTRODUCTION
Pan sharpening is a type of RS image fusion. The goal is to
obtain a high-resolution multi-spectral image, which combines
the spectral characteristic of the low-resolution data with the
spatial resolution of the panchromatic image. As a result of the
demand for higher classification accuracy of groundcover and
extraction of tiny geographic features (road, building,
vegetation, etc) information from some very highly resolution
satellite imagery collected from the sensor such as quickbird,
ikonos,and so on, many pan-sharpening algorithms were
developed. DigitalGlobe has evaluated a number of pan
sharpening algorithms in detail, UNB (stands for University of
New Brunswick) Algorithm is DigitalGlobe’s default pan
sharpening algorithm.
The objectives of this study are the new algorithm
development for fusing multi-spectral low-resolution remote
sensing image with a more highly resolved panchromatic
image collected by Quickbird sensor (see Table 1). PSF was
considered in this process of building new pan-sharpening
algorithm model.
Pan
725nm(450-900nm)
Blue
479.5nm(450-520nm)
Image Bands
Green
546.5nm(520-600nm)
Red
654nm(630-690nm)
Near IR
814.5nm(760-900)
Spatial
Pan
61 cm (nadir) to 72 cm (25
off-nadir)
Resolution
MS
2.44 m (nadir) to 2.88 m (25
off-nadir)
Table 1. Characteristics of the Quickbird data
A multispectral Quickbird scene consists of four bands called
Blue, Green, Red, and NIR. The first three bands characterize
the ground cover in the visible wavelength range and the four
band the near infrared wavelength range. Figure 1 depicts the
spectral characteristic of all Quickbird bands.
Figure 1. Spectral characteristic of Quickbird (C. Padwick,
2004)
2. RESEARCH DATA
The main research data is a standard bundle of pan and ms
image scene data, which is downloaded form the
DigitalGlobe’s website. The website also provide bundles of
pan-sharpening scene data of same data, it is convenient to
compare the new pan-sharpening algorithm to other
