Title :
Band selection in hyperspectral images based on the Fisher Linear Discriminant classifier
Author :
Harari, Y. ; Bar-Yehuda, Z. ; Rotman, S.R.
Author_Institution :
Dept. of Electr. & Comput. Eng., Ben-Gurion Univ. of the Negev, Beer-Sheva, Israel
Abstract :
Hyperspectral data consists of three dimensional images; the third dimension is the spectral signature of each pixel. The question always arises: which bands should be used and how should they be weighted. In this paper, a band selection algorithm based on the Fisher´s Linear Discriminant classifier (FLD) is implemented. An initial segmentation of the image into two different classes (gas and background) is used as an input for the FLD. The combination of bands to be used is selected based on the FLD results.
Keywords :
eigenvalues and eigenfunctions; image classification; image segmentation; Fisher linear discriminant classifier; band selection; hyperspectral images; image segmentation; spectral signature; Correlation; Detection algorithms; Eigenvalues and eigenfunctions; Hyperspectral imaging; Image segmentation; Pixel; Signal to noise ratio;
Conference_Titel :
Electrical and Electronics Engineers in Israel (IEEEI), 2010 IEEE 26th Convention of
Conference_Location :
Eliat
Print_ISBN :
978-1-4244-8681-6
DOI :
10.1109/EEEI.2010.5661942
