INTRODUCTION 
Myperspectral data presents interesting 
opportunities and challenges to researchers, not 
the least of which is the problems associated 
with displaying and manipulating imagery with 
over 200 channels and 16 bit resolution. The FLI 
has 288 spectral bands from 0.43 to 0.805 
microns and the AVIR1S has 224 spectral 
channels from 0.4 to 2.5 microns. The AVIRIS 
data also can be obtained spectrally-resampled 
to 210 channels. Most commercial image 
processing packages do not support imaging 
spectrometery sensors at their full spectral and 
radiometeric resolution. At the Institute for 
Space and Terrestrial Science (ISTS) we have 
developed an Imaging Spectrometer Analysis 
Package for the quantitative analysis of full 
resolution imaging spectrometery data. This work 
grew from a cooperative project between York 
University and Moniteq Ltd. to study data from 
the Canadian Fluorescence Line Imager (FLI) 
(Miller et al., 1987). 
IMPLEMENTATION 
At ISTS, software has been written to analyze 
imaging spectrometer data from the FLI and 
from NASA’s Airborne Visible/Infrared Imaging 
Spectrometer (AVIRIS) at full spectral, spatial 
and radiometeric resolution. The only limitation 
being the availability of disk space to store the 
very large volume of data: 16 MB for a typical 
FLI full spectral scene and 120 MB for a full 
AVIRIS scene. The image is usually subset to a 
more manageable size for analysis. The Imaging 
Spectrometer software is integrated into PCI’s 
EASI/PACE image analysis system using the 
Software Toolbox (PCI, 1989a). Imaging 
spectrometery data is stored consistent with the 
PCIDSK format (PCI, 1989b). Currently the 
software runs on a Microvax II, Sun workstation 
(SUN 3/60) and a PC based image processing 
workstation. 
A modular approach was taken to the software 
development. Instead of independent systems for 
batch and interactive processing of the imagery, 
one library of fundamental subroutines was used 
(see Figure 1) to which all modules are linked. 
In the interactive mode, FORTRAN modules 
were strung together by an EASI procedure 
(command file). This allows new analysis 
modules to be incorporated easily and permits 
each module to be run independently if the need 
arises. In the batch mode one large FORTRAN 
programme is used that does all the processing 
tasks. 
I'he first step in the analysis procedure, whether 
it be batch or interactive, is to select and 
calibrate the data. A tablet (or mouse) is used to 
define a window of interest from the displayed 
image for interactive processing. The average 
digital value for the window is calculated. For 
batch processing an image database input 
window is specified and calculations are done on 
a pixel by pixel basis. The data must be 
calibrated to appropriate units and for this 
purpose the software allows several options. 
Calibration files are stored for spectral, 
radiometric and reflectance calibration. 
Conversion to radiance is done using radiance 
scale factors supplied by the sensor’s 
manufacturer. One useful option is to employ a 
table of extra-terrestrial solar irradiance values 
and the image geometry to convert the image to 
pseudo-reflectance values. An approximate 
atmospheric correction procedure can be applied 
if ground measurements of suitable reflectance 
targets are supplied and a 
digital-number-to-reflectance calibration is 
available. If a Digital Terrain Model (DTM) is 
available a correction for topography can be 
made. The analyst determines which calibration 
procedure is to be used. 
Interpretation algorithms or inverse models can 
now be applied to calibrated data. These 
algorithms can extract spectral and/or 
reflectance parameters such as reflectance ratios, 
watercolour parameters (Freemantle and Miller, 
1987) or vegetation red edge spectral parameters 
(Miller et al., 1990b). In the interactive mode the 
calculated parameters and spectral curves are 
displayed at the workstation terminal for the 
analyst to assess (see Figure 2). The 
EASI/PACE parameter file stores interpretation 
choices, for example the vegetation red edge 
spectral window, for use during batch processing 
of the image. During batch processing user 
selected calculated parameters are written back 
to a PCIDSK image database in 8 bit form. This 
image can be then analyzed using conventional 
methods such as thresholding, pseudo-colour, 
etc. The modular approach allows new 
interpretation, atmospheric correction or 
calibration models to be added to the software 
with a minimum of effort. 
APPLICATION 
The Imaging Spectrometery Analysis Package 
has been used to analyze AVIRIS and FLI data 
over vegetated targets (Miller et al. 1990, Rock 
et al., 1990). The red edge analysis can be 
applied to both the FLI and AVIRIS data (see