Image Processing Reference
In-Depth Information
Chapter 11
Conclusions and Directions for Future
Research
11.1 Introduction
This monograph is a dedicated text dealing with the problem of fusion of
hyperspectral data for visualization purposes. A hyperspectral image consists of
a few hundred spectral bands which together encompass the reflectance response of
the scene over a large bandwidth. One can think of hyperspectral image as a col-
lection of a large number of bands that are narrow-band and contiguous in design.
The visualization of this huge multidimensional data over a standard display device
is an important but quite a challenging task. In this monograph, we explain various
solutions to this problem in the form of visualization-oriented image fusion. Through
each of the four solutions, we explore different aspects of the fusion process while
the primary objective of visualization remains the same. None of the presented solu-
tions assume any priori knowledge of the scene or the imaging system. The fused
image gets generated solely from the input hyperspectral data without requiring any
external information. In the next section, we provide some concluding remarks on
the fusion techniques discussed in the monograph. We also discuss future directions
for research in this area in Sect. 11.3 .
11.2 Conclusions
The fusion techniques combine disparate information provided by the multiple
spectral bands into a single composite image which is more informative than any sin-
gle band in the data. The process of fusion operates over individual pixels in the data.
Thus, the techniques presented in the monograph can be categorized as the pixel-
based fusion schemes. The pixel-based fusion schemes typically generate the fused
image through a weighted linear combination of input spectral bands where the
choice of the fusion weights is the most critical part of the fusion algorithm. Hence,
the weight function defines an
α
-matte for the purpose of combining the input bands.
 
Search WWH ::




Custom Search