Image Processing Reference
In-Depth Information
Fig. 5.3 Results of Bayesian fusion of the moffett 2 image from the AVIRIS. a Grayscale fused
image, and b RGB fused image (©2013 Elsevier, Ref. [91])
green, and blue channels of the display or printing device to generate the RGB com-
posite output. Figure 5.2 b shows the RGB version of the fusion result for the urban
data. It should, however, be noted that the colors in this image have been assigned
by a pseudo-coloring scheme as described, and they may not have any resemblance
to their natural meaning.
We also provide the results of Bayesian fusion for the moffett 2 data from the
AVIRIS. These data depict a very large number of small objects such as urban
establishments, and hence, we are particularly interested in observing how these
objects are represented in the resultant fused image. Figure 5.3 a shows the grayscale
version of the result of the Bayesian fusion. The RGB version of the fusion output
is shown in Fig. 5.3 b where each of the R, G, and B bands are generated from fusion
of nearly 1
3-rd bands of the moffett 2 hyperspectral image.
5.8 Summary
This chapter presents a Bayesian solution for fusion of hyperspectral images where
the primary purpose of fusion is quick and efficient visualization of the scene contents.
The fusion methodology relies on the statistical model of image formation which
relates observed hyperspectral bands to the fused image through a first order approxi-
mation. This model uses a parameter called the sensor selectivity factor (
). We have
discussed an approach for the computation of these parameters based on some of the
characteristics of the hyperspectral data. A Bayesian framework has been discussed
for the estimation of the fused image using the MAP estimator where the TV norm-
based prior has been employed. The fusion technique being completely data-driven,
does not require any information related to the sensor device, and hence, it tends to
be generic and sensor or device independent. The use of Bayesian framework not
Search WWH ::

Custom Search