Image Processing Reference
In-Depth Information
(a)
(b)
(c)
(d)
Fig. 4.6
Results of bilateral filtering-based fusion of the moffett
2
data from the AVIRIS applied
over a subset of bands selected using the entropy-based approach.
a
-
d
show the results of fusion
for
κ
values 0, 0.35, 0.50 and 0.70, respectively (Fig.
4.6
c: © ACM 2010, Ref: [87])
the moffett
2
data of different cardinality by varying the threshold parameter
.We
consider the resultant image from fusion of the entire dataset as shown in Fig.
4.6
aas
the reference for the quality assessment purpose. The result of fusion of 65 selected
bands of the data as shown in Fig.
4.6
b reduces the computations nearly to the one-
third of the original. However, the corresponding fusion result does not produce
any visible artifacts, and gives a PSNR value of 41dB when compared against the
aforementioned reference. The result of fusion of 29 selected bands was found to
give 35dB of PSNR (Fig.
4.6
c). A set of 12 image bands was selected by choosing
κ
κ
to be 0.70. The resultant fused image is shown in Fig.
4.6
d which was found to give
a PSNR of 30dB for the fusion of approximately
1
/
10
-th of the data.
Along with visual analysis, we also need to evaluate the quality of the fused
images obtained from the selected subsets with respect to the fused image obtained
from the entire dataset in terms of their probability distribution functions (pdfs). We
measure the extent of the similarity between the histograms of the corresponding two
images. We provide this analysis using the Bhattacharyya distance which measures
the overlap between the histograms of the two images. As in the previous case,
