Image Processing Reference
In-Depth Information
(e)
(f)
(g)
Fig. 10.3
(continued)
However, as the technique attempts to preserve the overall intensity of the fused
image with reference to the input data, the result appears a bit dull. The output of the
optimization-based technique, too, provides a high amount of contrast, and proves to
be better for visual interpretation. The results of the PLF and the CMF techniques as
shown in Fig.
10.4
f-g, respectively, have also captured most of the details from the
data. However, these two results lack in contrast and sharpness, giving the images a
washed-out appearance.
An objective assessment of these results can be found in Table
10.5
. The numerical
values of the no-reference measures are in agreement with our visual observations.
All the presented techniques yield fused images with high values of variance, entropy,
and sharpness indicating their usefulness in terms of contrast, average information,
and the perceived sharpness, respectively. The highest values of the gradient (
g
), and
¯
2
) are observed for the optimization-based technique which reflects
its ability to provide visually appealing results even from the data containing less
features. However, none of the techniques guarantee nearly equal contributions from
the constituent hyperspectral bands as the values of fusion symmetry (FS) are quite
the variance (
σ
