1227 
STUDY ON DATA FUSION METHODS WITH OPTIMAL INFORMATION 
PRESERVATION BETWEEN SPECTRAL AND SPATIAL BASED ON HIGH 
RESOLUTION IMAGERY 
Wenbo Li a,b *, Qiuwen Zhang b 
a Institute of Intelligent Machines, Chinese Academy of Science, 230031 Hefei, China - 791wb@163.com 
b Hubei Key Lab of Digital Watershed Science and Technology, Huazhong University of Science and Technology, 
430074 Wuhan, China 
KEY WORDS: Image processing, Integration, Image understanding, Fusion, Geography, Method, IKONOS, Quickbird 
ABSTRACT: 
Data fusion methods are often employed to balance contradiction between spectral information and spatial information of remotely 
sensed images. The fused images formed by an effective data fusion method based on remotely sensed data should have both high 
spectral information preservation with Low Spatial Resolution (LSR) images and high spatial information preservation with High 
Spatial Resolution (HSR) image. In this paper, Modification Brovey Transform (МВТ) has been proposed so as to fuse either 
individual band LSR images of high resolution imageries. Three fusion methods, such as Smoothing Filter-based Intensity 
Modulation (SFIM), Discrete Wavelet Transform (DWT) and МВТ have been used to fuse Quickbird images and IKONOS images. 
Both spectral information preservation and spatial information preservation of the fused images generated by three fusion methods 
based on two kinds of high resolution imageries have been evaluated by qualitative visual interpretation and quantitative statistical 
analysis. The evaluation results confirm that the proposed method has optimal spatial information preservation with HSR image in 
both fusion results. Meanwhile, SFIM fusion method has optimal spectral information preservation with LSR images. 
1. INTRODUCTION 
At present, the spatial and spectral resolutions of remotely 
sensed sensors are highly correlated factors. In general, the 
sensor with a higher spatial resolution (broader spectral range) 
is often achieved at the cost of a lower spectral resolution and 
vice-versa. On the same satellite or airplane platform, a 
panchromatic sensor covers a broader spectral range, while a 
multi-spectral sensor covers a narrower spectral range. With the 
development of technology, many high resolution sensors, such 
as Quickbird and IKONOS, have been successfully launched. 
The sensors of Quickbird and IKONOS both can provide four 
bands multi-spectral (MS) images and one band panchromatic 
(PAN) image with resolution ration 1:4. Four bands MS images 
provide more spectral information than PAN image and PAN 
image provides higher spatial information compared with MS 
image. Both spectral information and spatial information of 
remotely sensed images are very important for most remote 
sensing application. To take advantage of the spectral 
information of MS images and high space information of PAN 
image, data fusion techniques are often employed to generate 
fused images with high spectral and spatial quality 
simultaneously. Data fusion based on remotely sensed images is 
called image fusion. However, during the processing of image 
fusion, some useful information (spectral and spatial) will be 
lost, so an effective image fusion method should have optimal 
information preservation between spectral and spatial. 
Many image fusion techniques have been proposed to fuse low 
spatial resolution (LSR) images, such as MS images, with high 
spatial resolution (HSR) image, such as PAN image. These 
image fusion techniques can be divided into three fusion 
strategies. One strategy is based on RGB space transform, such 
as intensity-hue-saturation (IHS) (Haydan et al, 1982) transform 
and Brovey transform (ВТ) (Gillespie et al,1987). These 
techniques have the limitation that only three bands LSR 
images are involved. The second strategy is based on spatial 
filters, such as high-pass filtering (HPT) (Chavez et al, 1988) 
transform, smoothing filter-based intensity modulation (SFIM) 
(Liu, 2000), and different wavelet transform (WT) (Zhou et 
al, 1998; Nunez et al, 1999; Nencini et al,2007). The techniques 
belonging to this strategy can be used to fuse either individual 
band LSR images, independently. The third strategy is based on 
simple algebraic operation, such as multiplication-transform 
(MT) (Li et al,2006). It should state that there are no obvious 
limits between three strategies. For instance, ВТ belongs to both 
the first and the third strategies, while HPT belongs to the 
second and the third strategies. Meanwhile image fusion 
techniques belonging to different strategies are often used 
together (Audicana et al, 2004; Chibani et al,2002). 
In this paper, Modification ВТ (МВТ) has been proposed so as 
to fuse either individual band LSR images of high resolution 
remotely sensed images independently, such as Quickbird or 
IKONOS. Three fusion techniques, including SFIM, WT and 
МВТ, have been employed to fuse LSR images with HSR 
image of Quickbird and IKONOS. Both spectral information 
and spatial information preservation of three fusion methods 
have been evaluated by qualitative visual interpretation and 
quantitative statistical analysis. 
* Corresponding author. Wenbo Li, 791wb@163.com