Data Fusion in Remote Sensing of Urban and Suburban Areas - Remote Sensing of Urban and Suburban Areas - page 199

converts the known details of image A into the inferred details of image B k (inferred
details, Step 4). Finally, MSM -1 from resolution n°2 to resolution n°1 performs the
synthesis of the image B* kh (Step 5).

11.4.3 Illustration Example

Ranchin et al. ( 2003 ) propose a set of examples of different implementations of the
ARSIS concept. In the proposed example, the selected MSM is implemented
through the Undecimated Wavelet Transform (UWT) (Dutilleux 1989 ), the IBSM
is implemented through the Model 2 (Ranchin and Wald 2000b ; Ranchin et al.
2003 ) and the HRIBSM is the implemented through the Identity model. The three
models combined are referred to as the UWT-M2 method.
An Ikonos dataset has been used to test the UWT-M2 method. This Ikonos data-
set is composed of a panchromatic image Pan at a spatial resolution of 1 m and of
4 multispectral MS images at a spatial resolution of 4 m. The spectral bands of the
images are detailed in Table 11.1 . The geographical area is the city of Hasselt in
Belgium. The images were acquired the 28th of April 2000 at 10:39 UTC. The
original data are delivered with a dynamic range of 11 bits (gray values). Table 11.1
reports the mean value and standard deviation of each band in gray values. It also
provides the correlation coefficient between each band and the panchromatic image
resampled at 4 m. The application of the methods leads to the synthesis of MS
images at the spatial resolution of 1 m with the same dynamic range in 11 bits
Figure 11.6 (left) shows an excerpt of the Pan image at 1 m. On the left, there is
a river, crossed by two bridges, with small boats and several barges. Along the
riverbank is a main street. Several cars are visible. This area is mainly an industrial
district with large buildings, surrounded by numerous trees. Note those features at
the center of the image that appear in white in the upper left corner. On the top right
is a stadium. Its lawn is partly degraded. Figure 11.6 (right) presents an excerpt of
the same area in the near infrared (NIR) band with a spatial resolution of 4 m.
Figure 11.7 shows an excerpt of the NIR image synthesized at 1 m. The differ-
ence between this image and the original image at 4 m is striking. More details are
visible and the visual interpretation of the scene is easier and much more accurate.
Even the boats can be distinguished. One may note that the details (high frequencies)

Table 11.1 Spectral bands, mean value and standard deviation of the panchromatic (PAN) and
multispectral images (in gray values). Correlation coeficient between the original spectral bands
and the Pan image resampled at 4 m
Pan Blue Green Red Near infrared

Spectral band 450-900 450-520 520-600 630-690 760-900

Mean value 352 328 339 249 467

Standard deviation 55 26 39 48 124

Correlation coefficient 1.00 0.271 0.409 0.323 0.875