International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
DN (x, y) = DN, (x, y) + (DN, (x, y) - DN, (>, y))
Therefore,
DN,(x, v) = DN,(x, y) (6)
From the equation (6), the average value for a group of 4x4
sub-pixels in the new image is equal to their correspondent
pixel value in the IRMSS image. Obviously, the new image has
the same spectral energy as the IRMSS image within one
original pixel of the IRMSS image. So that, the data fusion
image produced by this method can keep the spectral energ
balance and can preserve the fidelity to the lower resolution
image spectral properties. The new method can also be called as
Spectral Energy Balance method.
4. APPLICATION
Based on the introduction as above, data fusion methods at the
present can be divided into two methods: spectral component
substitution techniques and spatial domain techniques. The
former replaces a spectral component of the multi-spectral
images by an adjusted high resolution image. The latter adds
high resolution information from the high resolution image to
all the low resolution spectral bands. Obviously, the spatial
domain techniques will not keep the spectral energy balance
and the spectral fidelity to the lower resolution image will be
destroyed because of adding a variable value to every pixel of
the low resolution image. According to hue theory, any colour
is quantitatively defined in terms of three variables: hue,
intensity and saturation, which give a numerical description of
the spectral range, brightness and purity of a colour. These
three variables are independent of each other (Qi, 1996). HIS
method for data fusion is to replace an intensity of the three
Band colour composite image with a high resolution image.
Based on the hue theory, intensity is independent of hue and
saturation to some extent, and the hue and saturation will not be
affected by such substitution in HIS data fusion. PCA method is
to replace the first principal spectral component in principal
component analysis of the multi-spectral images with a high
resolution image. Because the first principal component takes
the most information in principal component analysis, and it is
dependent of the solar incident radiance and the spectral
properties of the surface materials, any other band image cannot
represent the first principal component either in information
content or spectral properties. Therefore, the HIS method is
better than the PCA method in preserving the spectral fidelity.
Preserving spectral fidelity method also belongs to the spectral
component substitution techniques and it replaces a low
resolution band with an adjusted high resolution image, i.e.,
every pixel brightness value in the high resolution image is
adjusted by its co-registered pixel in the low resolution image.
Therefore, the fusion results produced by HIS and PSF methods
will be compared, mainly in preserving the spectral fidelity.
CBERS images in Guizhou Province, China, taken on July 18,
2000 are used for test. Image data is a 2 grade image data,
supplied by China Centre for Resources Satellite Data and
Application. CCD B4 image was selected as a higher resolution
image and IIMSS B6,B7,B8 were selected as lower resolution
images. Original image size is 400x200 pixels for CCD image
and 100x50 pixels for IRMSS images.
Three gray scale images based on CBERS CCD B4, IRMSS B7
and their data fusion image are shown if Fig.1: image marked A
is CCD B4; image B is IRMSS B7; image C is data fusion
image of B4 and B7 by PSF method. Image B and C show
clearly that the fusion images have more spatial details than
IRMSS B7, and have a similar spectral information to B7; From
image A and C, the fusion image contains the similar spatial
details in B4 image, but image brightness distribution in image
B is definitely different from image C.
Figure 1. Gray scale images based on CBERS CCD B4, IRMSS
B7 and their data fusion image: image A is B4; image B is B7;
image C is data fusion image of B4 and B7 by PSF method
Three Band colour composite image can more clearly show the
difference between the original image and the data fusion image.
Image marked A in Fig. 2 is a three band colour composite
image of B7, B8, B6; image B is a data fusion image derived
from B7.B8,B6 and B4 by HIS method; image C is a dal
fusion image by PSF method. Compared with image A. both
image B and C had more spatial details, such as some small
water system, which indicates that the data fusion method can
be used to enhance image spatial details and improve image
spatial resolution for lower resolution image; image C had the
similar colour to image A, while image B is quite different from
image A in colour. Which confirms that PSF method can be
used for preserving the spectral fidelity of the lower resolution
image in data fusion processing, but HIS method cannot
Intei
——
Obv
spec
A
Figu
imag
imag
fusic
PSF
Imag
An i
Fig.
imag
the o
a spe
by P
of B.
and t
Is ve
that
frequ
frequ
Whic|
andi
The i
fusio
imag
PSF |
From
than
912