The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008 
1170 
vegetation areas appear brighter than pavement areas in the 
IKONOS or Quickbird PAN image, meanwhile in the other 
PAN images such as SPOT, IRS or ALI, vegetation areas 
appear darker than pavement areas. As a result, the usual fusion 
methods are rarely suitable for all data and the good fusion 
quality depends on the data type and operator’s experience. 
2.2 Spectral response of Sensors 
For most earth resource satellites which provide both PAN and 
MS bands, in ideal condition, all MS bands would be well 
separated and would cover exactly the same wavelengths as the 
PAN band. In addition, the measured energy in the PAN band 
can be obtained with the summation of corresponding MS 
bands theoretically. However, there area no sensors show such a 
situation. Take the sensors onboard of IKONOS for example, 
the theoretical and actual spectral responses are shown in 
Figure 1. 
[—■Pull-But I Pan Blue Green Red NIR 
[—■Pull-But I Pan Blue Green Red NIR 
"la) (b) 
Figure 1. Relative Spectral response of IKONOS* (a) 
Theoretical spectral response; (b) Actual spectral response. 
In fact, the measured energy in an individual channel is the sum 
of incoming radiation and relative spectral response: 
L k = ^L{X)R k (A)dA (1) 
where X is the wavelength, L k the in-band radiance, L(X) at- 
aperture spectral radiance and R k (X) the peak-normalized 
spectral response. Therefore, the energy in PAN band of 
IKONOS can be estimated by defining four weights as follows: 
Pan= w B B+w g G+w r R + w nir NIR+ (other) (2) 
where Pan, B, G, R, NIR represent the radiance of individual 
spectral bands, w B , w G , w R , w NIR are the weights of 
corresponding MS bands, and other considers for the influence 
of the spectral range which missing from MS bands but still 
covered with the PAN band. For other satellites listed in Table 
1, the energy in PAN band can be obtained in the same way, 
and a general equation would be written as: 
‘http://www.geoeye.com/products/imagery/ikonos/spectral.htm 
(accessed 22 Oct. 2006) 
Pan = w i MS i + {other) (3) 
where MS’, is the corresponding MS bands which covered with 
PAN band, and w, is the weights of band i. It is suitable for 
most of the satellites which provide both PAN and MS bands. 
3. OUTLINE OF PROPOSED METHOD 
3.1 Fast IHS image fusion method 
In recent years, a variety of image fusion methods have been 
developed. According to its efficiency and implementation, the 
IHS image fusion method is probably the most one. To quickly 
merge massive volumes of data, Tu et al (2004) have proposed 
a fast approach of IHS fusion to perform the fusion process with 
lower computational cost. In the fast IHS method, the fused 
image[F(R), F(G), F(B)] T can be obtained from the upsampled 
original image [R, G, B] T easily by using addition operation, 
which is expressed as follows: 
'P(R) 
1 -1/V2 1/V2* 
'1 + iI^-I) 
F(G) 
= 
1 -1/V2 -1/V2 
vl 
F(B) 
1 
O 
1 
v2 
"l -1/V2 
1/V2 ’ 
'1 + 8' 
R + 8 
1 -1/V2 
-1/V2 
vl 
= 
G + Ô 
1 V2 
0 
v2 
B + 8 
where 8 = Pan — I and I = (R + G + B) / 3 . For IKONOS 
data fusion, given the spectral range of PAN image, Tu el al 
solves the spectral distortion problem by including NIR band 
into 7, that is I = (R + G + B + NIR) / 4 . To further 
consider the spectral mismatching between PAN and MS bands, 
a simple spectral adjustment is presented to use 
I SA ~{R + 0.75 *G + 0.25 * B + NIR) / 3 to replace I. 
3.2 Proposed fusion method 
According to its fast computing capability for fusing images, 
the fast IHS fusion method is widely used for fusion purposes 
and some modified methods have been proposed too 
(Choi,2006; Gonzalez-Audicana et al,2006). However, the 
adjustment and modification are mostly proposed for IKONOS 
and Quickbird images. Taking the sensor spectral response into 
account, we present a new improved method based on the fast 
IHS transform. The improvements are in two parts: 
(1) Construction of the intensity component 
Considering the relationship between the relative spectral 
response of MS and PAN sensors which discussed in section 2, 
the intensity component is generated by combining the MS 
bands whose spectral ranges are overlapped by the spectral 
coverage of the PAN band, no matter what combinations of MS 
bands are being fused. We can define the intensity component 
(I) as: