The Spectral Image Processing System (SIPS) Interactive
REMOTE SENS. ENVIRON. 44:145-163 (1993)
The Spectral Image Processing System
(SIPS) Interactive Visualization and Analysis of Imaging Spectrometer Data
F. A. Kruse, *,t A. B. Lefkoff,* J. W. Boardman,*
K. B. Heidebrecht,* A. T. Shapiro, * P. J. Barloon,* and
A. F. H. Goetz *'~
• Center for the Study of Earth from Space (CSES), Cooperative Institute for Research in
Environmental Sciences (CIRES), University of Colorado, Boulder tDepartment of Geological Sciences, University of Colorado, Boulder
T h e Center for the Study of Earth from Space
(CSES) at the University of Colorado, Boulder, has developed a prototype interactive software system called the Spectral Image Processing System (SIPS) using IDL (the Interactive Data Language) on
UNIX-based workstations. SIPS is designed to take advantage of the combination of high spectral reso- lution and spatial data presentation unique to im- aging spectrometers. It streamlines analysis of these data by allowing scientists to rapidly interact with entire datasets. SIPS provides visualization tools for rapid exploratory analysis and numerical tools for quantitative modeling. The user interface is X-Windows-based, user friendly, and provides
"point and click" operation. SIPS is being used for multidisciplinary research concentrating on use of physically based analysis methods to enhance sci- entific results from imaging spectrometer data. The objective of this continuing effort is to develop
Address correspondence to F. A. Kruse, CSES / CIRES, Univ. of Colorado, Boulder, CO 80309-0449.
Received 25 January 1992; revised 31 October 1992. operational techniques for quantitative analysis of imaging spectrometer data and to make them avail- able to the scientific community prior to the launch of imaging spectrometer satellite systems such as the Earth Observing System (EOS) High Resolution
Imaging Spectrometer (HIRIS).
Maps of the distribution and composition of ma- terials on the Earth's surface are an important source of information for scientific investigations of resources, environment, and man-made change on our planet. During the late 1980s and early
1990s, imaging spectrometry has emerged as an exciting technology that provides the potential for rapidly producing both traditional surface-cover maps and new maps based on quantitative mea- surement of Earth-surface properties. Imaging spectrometers acquire images simultaneously in many narrow, contiguous spectral bands (Goetz et al., 1985). The data can be thought of as a "cube" of the dimensions #lines x #samples x
#bands. NASA's operational imaging spectrome-
0034-4257 / 93 / $6. O0
©Elsevier Science Publishing Co. Inc., 1993
655 Avenue of the Americas, New York, NY 10010
1 4 6 Kruse et al.
ter, the Airborne Visible/Infrared Imaging Spec- trometer (AVIRIS), presently acquires data in up to 224 spectral bands (Vane et al., 1993). Data collected by these instruments can be displayed and analyzed as either images or as detailed spec- tra; one spectrum for each picture element in the image. High spectral resolution reflectance spectra collected by imaging spectrometers allow direct identification and characterization of indi- vidual materials including minerals, vegetation, water, ice and snow (Goetz et al., 1985; Vane and Goetz, 1985; 1986; Vane 1987; 1988; Green,
1990; NASA, 1987).
The strength of imaging spectrometry lies in the simultaneous use of spatial and spectral infor- mation for integrated analysis. Previous software packages, the Spectral Analysis Manager (SPAM)
(Mazer et al., 1988) developed at the Jet Propul- sion Laboratory (JPL) and the Integrated Software for Imaging Spectrometers (ISIS) developed at the U.S. Geological Survey in Flagstaff (Torson,
1989) utilized this concept to permit interactive analysis of subsets of imaging spectrometer data.
While providing basic capabilities, these packages did not satisfy many of the user community's sci- entific requirements primarily because they did not provide utilities for preprocessing or calibra- tion, only allowed analysis of a small part of the image cube, and were hardware-specific. Despite promising results from AVIRIS data using these and other software packages in the geological sciences, terrestrial ecology, hydrology, and ocean- ography, scientists have not yet tapped the full potential of the data. The volume and complexity of information contained in the imaging spectrom- eter data have made detailed analyses difficult, and only a small fraction of the data collected have ever been analyzed to extract quantitative information.
The Spectral Image Processing System (SIPS) is a software package developed by the Center for the Study of Earth from Space (CSES) at the
University of Colorado, Boulder, using IDL (the
Interactive Data Language, a proprietary pro- gramming language) (Research Systems Inc.,
1991) in response to a perceived need to provide integrated tools for analysis of imaging spectrome- ter data both spectrally and spatially. Many of the ideas and techniques incorporated into SIPS are the result of nearly 10 years experience with imaging spectrometer analysis by principals at
SIPS Version 1.1 H a r d w a r e Platforms a n d
R e q u i r e d Software Versions
Hardware Color System Manager IDL
DECstation 3100 S-bit Ultrix 4.1 Motif 1.1
DECstation 5000 8-bit Uhrix 4.1 Motif 1.1
IBM RISC 6000 8-bit AIX 3.1
Motif 1.1 IDL 2.2.2
SunOS 4.1 Openlook 2.0 IDL 2.2.2
SGI with xterm 8-bit IRIX 4.0 4Dwm IDL 2.2.2
CSES (Goetz, 1981; 1984; Goetz et al., 1985;
Goetz and Calvin, 1987; Goetz and Boardman,
1989; Gao and Goetz, 1990; Goetz and Davis,
1991; Kruse et al., 1985; 1990; Kruse, 1987, 1988;
Kruse and Dietz, 1991; Boardman, 1989; 1990;
1991). This manuscript describes the version 1.1 implementation.
SIPS was specifically designed to deal with data from AVIRIS and the High Resolution Im- aging Spectrometer (HIRIS) (NASA, 1987), but has been tested with other data sets including the
Geophysical and Environmental Research Im- aging Spectrometer (GERIS), GEOSCAN images, and Landsat TM. It takes advantage of high speed disk access and fast processors running under the UNIX operating system (Table 1) to provide interactive analysis of entire imaging spectrome- ter data sets. SIPS is specifically designed to allow analysis of single or multiple imaging spectrome- ter data segments at full spatial and spectral reso- lution. It also allows visualization and interaction analysis of image cubes derived from quantitative analysis procedures such as absorption band char- acterization and spectral unmixing.
SIPS version 1.1 presently consists of three modules: SIPS Utilities, SIPSView, and SIPS
Analysis (Fig. 1). The SIPS Utilities are programs for disk-to-disk processing of imaging spectrome- try data and include tape reading, data formatting, calibration to reflectance, and cosmetic process- ing of the data. SIPS_View provides interactive visualization and analysis capabilities for large imaging spectrometer data sets. It provides a user-friendly interface through the use of the
X-Window system and "widgets" such as menus, buttons, and slider-bars. SIPS_View provides the capability to interactively select and enhance bands to make color-composite images. It allows rapid extraction and display of individual spectra, or of spectra extracted from polygon regions.
These spectra can be visually compared to library
Spectral Imaging Processing System
s i p s _ u t i l s i p s
s i p s _ v i e w
s i p s _ a n a l
- c o n v e r t
- c v t 2 s i p s
- c v t 2 t e r r a _ m a r
- d n 2 r e f
- f i a t _ f i e l d
- i a r c a l i b r a t e
- m a k e _ b b l
- m a k e _ g a i n o f f
- m a k e _ h i s t
- m a k e s i p s _ c u b e
- m a k e _ s l b
- r d a v i m a g e
- rd a v w a v e
r e a d h e a d e r
- v i c a r i n f o
- i m a g e d i s p l a y
- c o n t r a s t e n h a n c e m e n t
- s p e c t r a b r o w s i n g
- s p e c t r u m
- v i e w s p e c t r a ( w / l i b r a r i e s )
- s p e c t r a l s l i c e
- s p e c t r a l m a t c h i n g
- u n n u x
- u n c o n s t r a i n e d
- p a r t i a l l y
- f u l l y c o n s t r a i n e d
Tree diagram showing functions of the Spectral
Image Processing System (SIPS). Brief descriptions of the
SIPS Utility functions are given in Table 2.
• Data visualization tools should be pro- vided for rapid, exploratory analysis.
• Numerical tools should be provided for quantitative modeling with the results dis- played visually in real time.
• The tools and techniques provided should be generally useful across multiple disci- plines.
• The software should have a user-friendly interface.
• The software should be independent of specific image display hardware.
SIPS UTILITIES spectra or automatically matched to spectral end- members. Extraction of spectral slices also allows display of spectral data as stacked, color-coded spectral images (Marsh and McKeon, 1983). SIPS
Analysis is a set of programs for detailed, full-cube analysis of imaging spectrometry data. These are primarily programs that require extensive mathe- matical calculations and CPU time that are not amenable to interactive analysis on a complete data set. Together, these three modules provide the capabilities to proceed from raw radiance data to final analysis results and output.
SIPS D E S I G N CRITERIA
The following requirements for the next genera- tion of imaging spectrometer software were de- fined based on an informal user survey and docu- mented research needs of CSES scientists in a variety of disciplines.
• The system should allow routine analysis of imaging spectrometer data sets to mini- mally include AVIRIS, GERIS, and Eos
• It should be flexible enough to permit lim- ited analysis of other multispectral data sets such as Landsat MSS, Landsat TM, and SPOT.
• The system should provide utilities for in- put of data, data formatting, data calibra- tion, and other common image processing tasks.
The SIPS utilities module contains tools that pre- pare data for input to SIPS_View, the analysis programs, and other image processing software.
These tools are written in IDL, with the exception of the tape reading utilities. The tape utilities are either IDL programs that spawn processes written in the C programming language or are written entirely in C. SIPS utilities operate on images with standard headers conforming to the
Planetary Data System (PDS) format (Jet Propul- sion Laboratory, 1991). The utilities all have a command line interface and some also have an interactive graphical interface. A list and brief description of tools presently available is given in
Table 2. Details of the header format including any variation from the PDS format and a complete list of parameters and detailed usage instructions for each tool are given in the
SIPS User's Guide
Most of the SIPS utilities format image data or create files for input to SIPS_View. For example, when starting with AVIRIS data on tape, the
utility is used to create a band sequen- tial (BSQ) and / or band interleaved by pixel (BIP) a n d / o r band interleaved by line (BIL) cube from the raw radiance data on that tape. The
utility' is used to create a wavelength table from the tape. To derive apparent reflectance values from the raw radiance values using the empirical line calibration (Roberts et al., 1985), the
make_ gainoffand dn2refutilities
are executed using the
BSQ cube as input. The data set for input into
SIPS_View is completed using
to create the calibrated BIP image cube, a histogram file and a bad-band-list file.
148 Kruse et al.
D e s c r i p t i o n o f S I P S U t i l i t i e s rd_avimage: rd avwave: vicar info: convert: cvt2sips: cvt2terra mar: make bbl: make hist: make_sips_cube: make slb: read header: rotate cube: subset_sips: make_gainoff: dn2ref: flat field:
Reads an image cube, with or without VICAR labels, from an AVIRIS tape; all present JPL AVIRIS tape formats are supported
Reads the wavelength and F W H M data from an AVIRIS tape and outputs wavelength file; this wavelength file is used as input to SIPS View
Displays the VICA-R label information for each tape file
Converts the storage order of an input cube that is in BSQ, BIP, or BIL to either BSQ, BIP or BIL format
Creates an output cube with a standard SIPS h e a d e r from BSQ, BIP, or BIL formatted non-SIPS data with any size header
Creates an output cube with a standard Terra-Mar header (used in Terra-Mar M I C R O I M A G E software) from a
BSQ input cube
Creates a bad-bands file; the bad-bands file can be used as input to SIPS_View to mask out bad bands during spectral processing
Creates a histogram file from an input BSQ cube file with a standard SIPS header; the output file is used as input to SIPS View for rapid contrast stretching
Creates a B S Q formatted output cube with a standard SIPS header from multiple non-SIPS input image files; the output cube can be used as input to most other utilities and to SIPS_View
Creates a SIPS spectral library file containing any n u m b e r of ASCII spectra files
Looks for a standard SIPS header in t h e input file and prints out the header information; if no h e a d e r is found, it prints an error
Rotates the cube 90 °, 180 °, or 270°; this utility can b e used to change the orientation of the image in the Image W i n d o w
Extracts a subset of an input cube and writes it to a cube file with a standard SIPS header
Uses the empirical line m e t h o d using selected areas on the g r o u n d to calibrate the data to reflectance (Roberts et al.,
1985; Elvidge, 1988; Kruse et al., 1990). This calibration m e t h o d requires choosing two or m o r e ground target regions with diverse albedos and acquiring field or laboratory spectra to characterize them.
uses the ground target reflectance spectra and the associated image radiance spectra to perform a linear regression for each band to d e t e r m i n e the gains and offsets required to convert the DN values to reflectance. The output is a file containing the gains and offsets that can be used as input to
Applies gains and offsets calculated by
to the entire imaging spectrometer cube; it uses a raw radiance image file and outputs a calibrated apparent reflectance cube
Removes a single s p e c t r u m for an area selected by the user with a uniform spectral response from the entire cube by division; it uses a raw radiance image file and outputs a calibrated apparent reflectance cube
Removes the global average spectrum for a cube from the entire c u b e by division to calibrate to Internal Average Relative
Reflectance (IARR) (Kruse, 1988); it uses a raw radiance image file and outputs a calibrated apparent reflectance cube
The remainder of the SIPS utilities operate on image data to prepare them for input to various image processing software. For example,
create an image file with a standard SIPS header. As another example, the
utility converts an image file with any type of header to an image file that can be used with the Terra-Mar "Microlmage ®" software
SIPS_View is an interactive IDL program that allows the user to visualize and work with imaging spectrometer data both spectrally and spatially. It uses "widgets" along with mouse and keyboard input to create a user-friendly interface. A widget is a simple graphical object, such as a pushbut- ton, slider, or menu that allows users easy interac- tion with the program. A more detailed descrip- tion of widgets is given in Table 3. Interaction by the user on a given widget produces what is re- ferred to as an "event" from that widget. When the user generates an event (by pushing a button, moving a slider, etc.), the software is then able to respond to the event by performing some func- tion. An example of the SIPS_View main window using IDL widgets under Motif (OSF, 1989) is shown in Figure 2.
In addition to the main window, SIPS View creates and manages many other windows
throughout its execution. The "Status Window" is a normal text widget on the lower right hand corner of the SIPS View main window that dis-
Spectral Imaging Processing System 149
Table 3. Definition of Widget Types and Actions
M e n u widget
Used to select a given option. It consists of a rectangular region with a label. "Pushing" the button by moving the m o u s e cursor over the button and pressing the left m o u s e button generates an event
Used to select a value from a range of possible values. It consists of a rectangular region in- side of which is a sliding pointer that displays the current value. The slider is "grabbed" by placing the m o u s e cursor over the slider pointer and holding down the left m o u s e button. Moving the m o u s e while continuing to hold the m o u s e button down will change the slider value
Used to select one or more items from a given list. It consists of a rectangular region with a list of items, each with its own toggle button with an "on" and "off" state
M e n u widget
Used to select a given option from a list of options.
It consists of a button widget that when pressed expands into the list of choices. T h e m e n u m a y be viewed (or the m e n u item executed) by moving the m o u s e cursor into the m e n u button and clicking the left m o u s e button
Used to select one item from a list of items.
It consists of a rectangular region with a list of items, one item per line. Moving the m o u s e cursor over an item and clicking the left m o u s e button generates an event and selects the item
Used to receive user input from the keyboard.
It consists of a rectangular region that when
"'activated" will display and act on characters typed from the keyboard
Used to display text in a window. It consists of a rectangular region, usually s u r r o u n d e d by a frame, w h e r e the program may display text
Used to display a standard I D L graphics window within a widget application. It consists of a rectangular region w h e r e plots and images are displayed plays useful information about the current state of S I P S V i e w , and the functions that the three mouse buttons will currently perform. The last two lines of the Status Window report processing status. The "Scroll Window" (not shown in Fig.
2) displays a subsampled image if a data set is larger than the standard 512 line × 614 pixel
AVIRIS image and allows extraction of a full reso- lution image for a desired location. The "Image
Window" contains the image at full resolution.
The "Zoom Window" contains a subset of the image zoomed from 1 to 16 times. The "Current
Spectrum Window" and "Saved Spectra Window" are used for viewing, extracting and saving spec- tra. Other windows such as "View Spectra," "Spec- tral Slices," and the "SAM Viewer" are created only when accessed by the menu fractions.
SIPS_View requires as a minimum one image cube in either BSQ, BIL, or BIP format to run.
Optimum functionality and performance are ob- tained if both a BSQ and BIP file are present.
The second copy of the data in BIP format allows quick access to individual spectra. Other optional files including a wavelength file, histogram file, bad-band file, and spectral libraries enhance the performance and utility of the program. The wavelength file allows the SIPS_View to associate each band to a specific wavelength. The histogram file allows the program to quickly perform stretches of 16-bit data to byte data for display.
The band-band file allows users to mask out un- wanted bands when plotting and analyzing spec- tra. File characteristics and formats are described in detail in the SIPS User's Guide (CSES, 1992).
SIPS_View Display Functions
Nearly all of the available functions in SIPS_View perform their operations in the Image, Zoom,
Current Spectrum, and Saved Spectra Windows located within the main SIPS_View window (Fig.
2). The display functions operate on the image data in its spatial format and are accessed by clicking on the pull-down "Image Menu." When
SIPS_View is started, a gray-scale image is dis- played. If the image is larger than 512 × B14,
SIPS_View will display the full image subsampled to fit into the Scroll Window. Scrolling is used to allow the user to view different portions of the image at full resolution. The Image Window dis- plays the full resolution image in a 512 line ×
614 sample window with a default 2% linear contrast stretch applied. The displayed image can be a gray-scale or density-sliced image of a specific band, a color composite image of three bands, or a gray-scale image of analysis results. An example of a typical SIPS user session screen is shown in
Figure 3. Possible actions associated with the
Image Window include:
• selecting which band is displayed
• selecting the display mode, zoom factor, and contrast stretch of the band displayed
150 Kruse et al.
The initial view of the SIPS_View main window using IDL widgets under Motif.
• saving the current image to a data file or a color PostScript file
Individual bands can be selected by entering the band n u m b e r or wavelength, or by using slider bars (Fig. 3). Grabbing and moving the slider changes its value and thus the band displayed.
SIPS_View can display the current band as a gray-scale or a density-sliced image, or display three bands as an RGB color composite. Selecting
"Toggle Color On" on the pull-down Image Menu causes S I P S V i e w to replace the single slider used for selecting a single band n u m b e r with three sliders used for selecting a red, green, and blue band number. Grabbing any of these sliders and selecting a new band n u m b e r to display in that color causes SIPS_View to read and display the new color composite image when the slider is released. Selecting "density-slice" on the m e n u provides an 18-color RGB ramp for a single image where low brightness values are represented by blacks and blues and high brightness values are represented by reds and whites. Individual ranges
Spectral Imaging Processing System 151
Typical SIPS user-session screen showing some of the fimctions used for spectral analysis of imaging spectrom- eter data. can also be manually selected for each color of the density slice. the current pixel are displayed under the Zoom
The Zoom Window
The Zoom Window, located to the right of the
Image Window (Figs. 2 and 3) displays a small area from the Image Window with a user-defin- able zoom factor applied. The center of the zoomed area is defined by the position of the cursor in the Image Window. The pixels displayed can be magnified from 1 to 16 times their original size by grabbing the slider titled "Zoom Factor" and changing its value to the desired new zoom factor. The positions of the four red corner indica- tors in the Image Window change to reflect the new area displayed in the Zoom Window. The line and sample coordinates and data value of
Interactive Contrast Enhancement
The user can change the contrast stretch for each band displayed by altering how the data fit into the 0 to 255 (8-bit) display range. The "Contrast
Stretch" option is selected via the pull down Im- age Menu button. SIPS_View creates a new win- dow for the e n h a n c e m e n t functions (Fig. 4). This window contains three draw widgets and plots the current histograms of the red, green, and blue bands. W h e n the current image is a gray scale or density slice, only the first window is used. Next to each plot, SIPS_View displays the band num- ber, and the m i n i m u m and maximum values used for the current stretch. If there is a histogram file
1 5 2 Kruse et al.
SIPS Histogram Window showing options for interactive contrast stretching of imaging spectrometer data. associated with the input image, then a slider is also provided that allows the user to choose a specific percent of the data to stretch (0-15%).
Two vertical bars within the plot display graphi- cally where the current m i n i m u m and maximum values lie on the histogram. W h e n e v e r a change is made to the m i n i m u m and maximum stretch values, the position of the two vertical lines in the
Spectral Imaging Processing System 153
histogram plot changes as well. The currently displayed image in the Image Window will not be updated with the new stretch, however, until one of the three "Apply Stretch" buttons is se- lected. When applied, all values less than or equal to the minimum stretch value are set equal to 0, and all values greater than or equal to maximum stretch value are set equal to 255. All values be- tween the minimum and maximum are
linearly into the 0-255 range.
The displayed image can be saved to a file as byte- scaled images of either a single-band, gray-scale image or a three-band, RGB image with the cur- rent color tables applied. The output files are in
BSQ format with the first band corresponding to the red values, the second band to the green values, and the third band to the blue values. This allows easy data interchange and also provides contrast enhanced images for use with color film writers or other output devices. SIPS_View also gives the user the option of saving files in color
SIPS_View Spectral Functions
SIPS_View spectral functions are those items that deal with imaging spectrometer data primarily in its spectral format. These functions include spec- tra browsing, spectra averaging, spectral slice ex- traction, viewing spectra (with spectral libraries), and Spectral Angle Mapping (SAM).
SIPS View allows the user to move the mouse cursor around the Image Window displaying the current spectrum in real-time. If SIPS_View is being executed with either a BIL or BIP cube, then the spectrum at the current cursor position is displayed in the Current Spectrum Window and continuously updated as the cursor is moved using the mouse. Alternatively, if only a BSQ cube is present, then a click of the left mouse button causes the spectrum at the current cursor position to be extracted band by band and displayed in the Current Spectrum Window (a process that takes 5-15 s per spectrum on a DECstation 5000).
At any time while browsing, a spectrum may be saved into the Saved Spectra Window by clicking the middle mouse button. Once a spectrum is in the Saved Spectra Window, it may be saved to disk in an ASCII or binary library format, further examined in the View Spectra Window, or used as input for analyses.
This function allows the user to interactively de- fine and extract spectra for irregularly shaped polygon regions, vectors, and individual pixels.
SIPS_View provides five totally independent classes (Fig. 2). Any one class can contain up to
10,000 pixels from one or more separately defined regions. Only one class is "active" at a time, and the statistics for that class are totally independent of the other four classes. For example, ifa polygon is defined when Class 1 is active, the spectra of
pixels contained within this polygon region will be averaged only with other regions defined for Class 1.
The region of interest is selected by position- ing the Zoom Window coverage and setting the zoom factor. Selecting the "define class" button creates a new window with the same spatial cover- age as the Zoom Window. Classes are defined by
"drawing" polygons and / or vectors and / or select- ing individual pixels within the selection window.
After the class regions are defined, SIPS_View extracts the spectra for all the pixels contained within the regions and calculates the mean, stan- dard deviation, and minimum and maximum spec- tra. The results of the calculations are plotted in the Saved Spectra Window. This process may be repeated any number of times on separately defined regions in the image. Every time a new region is defined and extracted, the computed statistics for that region are averaged with the overall statistics for all the regions previously de- fined for that class. For each individual class, the control panel shows the class name and the total number of pixels contained in all the polygon regions defined for that class. The user can toggle between showing statistically-derived spectra for a single class or showing the mean spectra of all classes in the Saved Spectra Window. SIPS_View also allows saving to disk files (as either ASCII spectra or binary spectral library files) the user's choice of the following:
• the coordinates of all of the pixels in a given class
• the mean spectrum
154 Kruse et al.
Figure 5. SIPS View Spectra Window showing three laboratory spectra (illite (ill07.usg), dolomite (cod2005.usg), and calcite (co2004.usg)) and three spectra extracted from an imaging spectrometer cube for an area with sericite (muscovite or illite) (pixel 481,366), calcite (pixel 424, 217), and dolomite (pixel 424, 220). The laboratory spectra are resampled to
AVIRIS resolution. Windows for interactive selection of both ASCII and binary libraries are an integral part of this utility.
• one standard deviation spectrum above the mean
• one standard deviation spectrum below the mean
• spectrum of the cumulative maximum at each wavelength
• spectrum of the cumulative m i n i m u m at each wavelength
"View Spectra" is a utility used for spectral display and analysis. W h e n the View Spectra function is selected, SIPS_View creates a separate window to plot the spectra currently in the Saved Spectra
Window as well as access and plot other ASCII and library spectra saved in binary format (Fig.
5). The user can then manipulate this plot in a n u m b e r of different ways, produce a PostScript output file of the plot, or import the plotted spectra back into the Saved Spectra Window for subsequent use in other SIPS functions.
The plot within the View Spectra Window initially displays the spectra that are in the Saved
Spectra Window. Image spectra from the Saved
Spectra Window are divided by a scale factor of
1000 (the same scaling factor used by SIPS Utili-
Spectral Imaging Processing System 155
ties to preserve precision in reflectance calibra- tion) so that both the image spectra and the laboratory spectra are the same magnitude (0-1.0) when plotted. SIPS_View plots each spectrum in a different color (up to 16 colors), and the names of the displayed spectra are listed in matching color on the right side of the plot. If a bad-bands file is present and the bad-bands filter is on, then all spectra containing the same number of bands as the image data will be plotted with only the good bands showing. Any spectrum containing a different number of bands will ignore the bad- bands list.
The two areas in the left column below the plot titled "Library Spectra Files" and "ASCII
Spectra Files" are used to select library files and
ASCII spectra files to plot (Fig. 5). Both areas contain an editable text widget titled "Path:" and a list widget below listing the matched files of the path. Moving the mouse cursor over the desired file name in the ASCII Spectra Files list and clicking the left mouse button causes the selection to be highlighted and the spectrum to be read from disk and plotted. Moving the mouse cursor over the desired file name in the Library Spectra
Files list and clicking the left mouse button high- lights the selection and opens the library file.
SIPS_View displays this library as the Current
Library, and lists the first 256 elements of this library in the middle column. Moving the mouse cursor over the desired element name in the list and clicking the left mouse button select and plots this spectrum with any spectra already plotted.
The View Spectra Window provides various mechanisms for manipulating the spectra once they have been plotted. The scale of the plot is automatically set to include all the spectra se- lected each time a new spectrum is selected. The user may also explicitly change the scale either by entering starting and ending wavelengths or reflectance values, or by clicking on the appro- priate axes. SIPS_View can either plot the spectra in the View Spectra Window overlaying one an- other, or stacked vertically offset from one an- other. Stacking the plot is useful for comparing spectra that have similar shapes and reflectance values (Fig. 5).
SIPS_View can apply a fast Fourier transform filter to any plotted spectrum, allowing smoothing of noisy data. The FFT filter is used by grabbing the slider titled "FFT Filter" and changing the value. Upon releasing the slider, the spectra are replotted with the filter applied. The higher the value of the slider, the fewer harmonics are used to draw the spectra, and, thus, the "smoother" the spectra appear. If there are very noisy bands in the data and these bands are not masked with a bad-bands list, then attempting to smooth the spectra using the FFT function may result in harmonic "ringing." Also, filtering will only affect spectra with the same number of bands as the image data.
Color PostScript output can be produced of the View Spectra plot exactly as it appears. If the plotted spectra are stacked and smoothed with the FFT Filter, then that is how the plot will be saved to the PostScript file.
SIPS_View spectral libraries are binary files that contain spectra and their associated wavelengths
SIPS User's Guide
(CSES, 1992) for format information]. They also have an associated auxil- iary information file. SIPS includes two sets of libraries of laboratory spectra. Digital spectra for approximately 135 minerals are provided courtesy of Jet Propulsion Laboratory (Grove et al., 1992).
The second set consists of digital spectra of 25 well characterized minerals, each measured on five different spectrometers as part of Interna- tional Geologic Correlation Project 264 (IGCP-
264, Remote Sensing Spectral Properties, Kruse, unpublished data).
The JPL spectra are hemispherical reflec- tance measurements from 0.4/tm to 2.5/lm made on a Beckman UV5240 spectrophotometer. The sampling interval is every 0.001 /tm (1 nm) be- tween 0.4/~m and 0.8 pm and 0.004/~m (4 nm) from 0.8 /~m to 2.5 /tin. Spectral resolution is approximately 1% of the wavelength measured.
Two sets of three (six total) spectral libraries are provided corresponding to three grain sizes (125-
500/tm, 4 5 - 1 2 5 / l m , and < 45/~m) measured at
JPL. One set of the three libraries is provided at full resolution to allow use of this resolution or resampling to specific AVIRIS wavelengths. The second set of the three libraries is provided resam- pled to 1989 AVIRIS wavelengths (data prior to
20 September 1989).
The IGCP-264 spectral libraries included in
SIPS represent the prototype database of approxi- mately 25 well-characterized minerals identified
156 Kruse et al.
as critical for geologic mapping by a 1987 inter- national survey (Kruse, unpublished data). Five spectral libraries measured on five different spec- trometers for the 25 minerals are provided. The same samples were measured on a Beckman
UV5270 spectrophotometer at CSES, a Beckman
UV5240 spectrophotometer at the U.S. Geologi- cal Survey in Denver (Clark et al., 1990), on the "RELAB" spectrometer at Brown University
(Pieters, 1990), on the "SIRIS" field spectrometer in the laboratory at CSES (Geophysical and Envi- ronmental Research, 1988), and with the proto- type of a new high resolution field spectrometer, the "PIMA II" (manufactured by Integrated Spec- tronics Pty. Ltd.), in the laboratory at CSES. The
CSES Beckman lab spectrometer measures at constant 3.8 nm resolution (sampled at 1 nm) throughout the 0.7-2.5 /~m range. The USGS spectra are provided at the standard "1 ×" resolu- tion ranging from 2 nm to 10 nm in the 0.4-2.4
/~m range and falling off to nearly 30 mn in the
2.4-2.5 nm range. The RELAB spectra are pro- vided at 2-13 nm resolution (sampled at 5 nm) in the 0.4-2.5/,tm range. The SIRIS spectra are sampled from 2 nm to 5 nm in the 0.4-2.5/~m range; however, spectral resolution information for this specific instrument has not been deter- mined. The PIMA spectra are sampled at 2 nm in the 1.3-2.5/~m range. While the spectral reso- lution function is not yet available for this instru- ment, comparison to measurements from the other spectrometers indicates that resolution is better than 4 nm throughout the measured spec- tral range. All spectra were measured with halon as the reference and reduced to absolute re- flectance using a NBS halon spectrum. A sixth
IGCP-264 spectral library consists of the library spectra measured on the USGS spectrometer re- sampled to 1989 AVIRIS wavelengths (data prior to 20 September 1989).
Spectral Slices (Stacked, Color-Coded Spectra)
Extraction of spectral slices from the images allows display of spectral data as color-coded stacked spectra (Marsh and McKeon, 1983; Kruse et al., 1985; Huntington et al., 1986) (Fig. 3). The color slice uses a standard 18-level density slice where white and red correspond to high intensity values and black and blue correspond to low intensity values. SIPS_View is able to extract three different types of slices: horizontal, vertical, and arbitrary. A horizontal slice extracts spectra along a horizontal line in the image. A vertical slice extracts spectra along a vertical line in the image (Fig. 3). An arbitrary slice extracts spectra along an arbitrary, user defined path. Each ex- tracted slice occupies its own window. Up to five slice windows may be open at any one time.
There are two general stages to using the slice option. First, SIPS_View extracts the desired slices the user has defined and places them into their own windows. Once these windows have been created, the user can move the mouse cursor around in the slice windows and see a specific spectrum plotted in the Current Spectrum Win- dow. The line and sample position for the pixel associated with that spectrum are listed in the
Slice Window, as well as the band and reflectance value under the cursor. The Zoom Window also follows along and updates the current pixel loca- tion.
The Spectral Angle Mapper (SAM)
The Spectral Angle Mapper (SAM) is a tool that permits rapid mapping of the spectral similarity of image spectra to reference spectra (Boardman,
1993a). The reference spectra can be either labo- ratory or field spectra or extracted from the image.
This method assumes that the data have been reduced to "apparent reflectance," with all dark current and path radiance biases removed. The algorithm determines the spectral similarity be- tween two spectra by calculating the "angle" be- tween the two spectra, treating them as vectors in a space with dimensionality equal to the num- ber of bands (nb). A simplified explanation of this can be given by considering a reference spectrum and a test spectrum from two-band data repre- sented on a two-dimensional plot as two points
(Fig. 6). The lines connecting each spectrum- point and the origin contain all possible positions for that material, corresponding to the range of possible illuminations. Poorly illuminated pixels will fall closer to the origin (the dark point) than pixels with the same spectral signature but greater illumination. Notice, however, that the angle be- tween the vectors is the same regardless of their length. The SAM algorithm (Boardman, 1993a) generalizes this geometric interpretation to nb- dimensional space. The calculation consists of taking the arccosine of the dot product of the spectra. SAM determines the similarity of a test
Spectral Imaging Processing System 157
Band2 / / spectrum
6. Plot of a reference spectrum and test spectrum tbr a two-band image. The same materials with varying illu- mination are represented by the vectors connecting the or- igin (no illumination) and projected through the points rep- resenting the actual spectra. spectrum t to a reference spectrum r by applying the following equation: c o s
\11£111" I1 11/ which can also be written as ub \
Et, r, I
/ where nb = n u m b e r of bands.
This measure of similarity is insensitive to gain factors because the angle between two vectors is invariant with respect to the lengths of the vec- tors. As a result, laboratory spectra can be directly compared to remotely sensed apparent reflec- tance spectra, which inherently have an unknown gain factor related to topographic illumination effects.
SIPS_View allows up to 10 reference spectra to be processed simultaneously using SAM. Spec- tra can be selected from SIPS spectral libraries,
ASCII spectra files, or any spectra contained in the Saved Spectra Window. Reference spectra must have the same wavelength set as the image to which they will be compared. If a bad-bands file is associated with the image cube, then SAM will use the reference spectra with the bad-bands masked, ignoring the bad bands in the calcula- tions. Optionally, all bands can be included, and
SAM will perform its calculations on all the bands over the defined range.
For each reference spectrum chosen, the spec- tral angle a is d e t e r m i n e d for every image spec- trum, and this value, in radians, is assigned to that pixel in the output SAM image. A unique spectral range may be chosen for each reference spectrum.
This allows the algorithm to focus on spectral regions that are significant for a particular refer- ence spectrum. The derived spectral angle maps form a n e w data cube with the n u m b e r of bands equal to the n u m b e r of reference spectra used in the mapping. Results can be viewed immediately using the interactive SAM Viewer (Fig. 7). The dynamic nature of the viewing interface helps the user to analyze the spatial patterns of spectral variability in the image and to rapidly map areas that are spectrally similar. The SAM Viewer Win- dow displays the results for each reference spec- trum separately as gray-scale images. Small spec- tral angles correspond to high similarity and these pixels are shown in the brighter gray levels.
Larger angles, corresponding to less similar spec- tral shapes, are shown in the darker gray levels.
Two interactive sliders, titled "Low Threshold" and "High Threshold," can be used to fine-tune the contrast stretch of the image. Values between the two slider settings are stretched linearly be- tween black and white. Values outside this range are set to black if they have a spectral angle greater than the High Threshold setting (less simi- lar to the reference) and white if they have a spectral angle less than the Low Threshold setting
(more similar to the reference). In addition, the two sliders can be "locked" one radian apart. In the locked setting, the image displayed is a binary map of all pixels more similar than the High
SIPS ANALYSIS PROGRAMS
The analysis module provides tools that perform complex calculations on an entire image and are too time-consuming for interactive use. Currently, only the
analysis tool which performs linear spectral unmixing is available in this module. A knowledge-based, expert system analysis utility is presently undergoing testing and revision (Kruse
1 5 8 Kruse et al.
SAM Viewer Window showing gray-scale results for comparison of image spectra to a reference spectrum corresponding to dolomite. Brighter areas represent better matches. et al., 1993) and will be released in the next version of SIPS. Other analysis modules are being developed and will be added at a later date.
Spectral mixing is a consequence of the mix- ing of materials having different spectral proper- ties within the GFOV of a single pixel (Singer,
1981; Smith and Adams, 1985; Boardman, 1991).
The SIPS unmixing program, written in IDL, uses a simple linear mixing model. This model assumes that observed spectra can be modeled as linear combinations ofendmembers contained in a spec- tral mixing library (Boardman, 1993b). The un- mixing approach seeks to determine the fractional abundance of each e n d m e m b e r within each pixel.
Given more bands than mixing endmembers, the problem can be east in terms of an overdeter- mined linear least-squares inversion for each im- age spectrum (Fig. 8) (Boardman, 1989; 1990;
The SIPS unmixing program provides three types of unmixing algorithms: unconstrained, par- tiaIly constrained, and fully constrained (Board- man, 1990). The unconstrained version provides a classic least-squares solution to the unmixing problem and the derived abundances are free to take on any value including negative ones. In the constrained versions the derived abundances are required to be nonnegative. When fully con- strained, their sum must be unity or less (Fig. 9).
Several steps are involved in using the unmix- ing program. Any user should become familiar with the concept of unmixing and its inherent assumptions and limitations before trying to un- mix their data (Singer, 1981; Smith and Adams,
1985; Boardman, 1991 and references therein).
The first and most important step is the selection of the spectral mixing library endmembers. In the
SIPS unmixing procedure, the mixing library is formed by interactively choosing members from the imaging spectrometer cube or from any num- ber of spectral libraries using SIPS_View. The unmixing library should contain all the materials believed to be mixing in the scene. Conversely, it should not contain members that are not present.
Once the mixing library is formed, the endmem- ber spectral are displayed, and a subset of the full spectral range can be chosen to ignore noisy bands and any spectral variability in wavelength
mixing endmernber library
Spectral Imaging Processing System 159
a b u n d a n c e s ) f
mixing endmember library
% endmember A
Linear spectral mixing forward and inverse mod- els. If the number of endmembers in the library is less than the number of bands in the data, then the problem is an overdetermined linear least squares inversion.
Sketch of the constrained inversion solution space for two endmembers. Best-fitting abundances must be positive and sum to unity or less. The reconstruction fit error must be greater than or equal to that for the uncon- strained solution. regions that are not of interest. Bad bands can also be excluded using the bad-bands file. Once the type of unmixing constraints are chosen, the program inverts the spectral library using singular value decomposition (Golub and Van Loan, 1983;
Press et al., 1986; Boardman, 1991; 1993b), and optionally displays the singular values of this li- brary matrix. The library's degeneracy is deter- mined by examining the products of the decompo- sition. If the library of endmembers consists of completely spectrally separable, orthogonal end- members, the normalized singular values will all be equal. For a wholly degenerate library, all but one singular value will be zero, indicating that all of the endmembers are linearly scaled versions of each other. At this stage, the user can choose to start again and revise the library. Finally, once the user is satisfied with the library selected, the program processes the full image data cube one line at a time. The unconstrained Unmix program runs in about 1 h for a standard AVIRIS scene while the fully constrained Unmix program takes about 5 h (times for five endmembers on a DEC-
Station 5000 / 200).
The output of the unmixing process is another image data cube. It has the same spatial dimen- sions as the input data. The number of output bands is equal to the number of endmembers plus two. This cube contains one image for each endmember showing the derived spatial patterns of abundance for that endmember. The additional two images are useful in assessing the uncertainty in the unmixing results. They are 1) an image of the sum of the abundances at each pixel and 2) the root-mean-square (RMS) error at each pixel.
The error image displays how well the mixing library can be used to model each observed spec- trum, and can be used to assess the validity of the mixing library. If contiguous regions of high error exist, a required mixing endmember was probably omitted. Refinement of the results involves itera- tire unmixing with revised libraries until the RMS errors are low. The resulting abundance images comprise estimates of the spatial distribution of the mixing e n d m e m b e r materials.
TYPICAL USER SCENARIO
The following is a typical SIPS user scenario for
AVIRIS data acquired for geologic investigations.
Many of the steps are common to analysis of any type of imaging spectrometer data and illustrate some of the relations between the different parts of SIPS. The scenario is presented sequentially
(in the normal order executed) in outline format
160 Kruse et al.
to clearly show the logical progression of the steps involved in the analysis.
receive and read AVIRIS tape using
getting BSQ im- age and wavelength files
• view radiance images using SIPS_View to verify data location and quality; locate and extract polygons for calibration areas
• build spectral library containing ground spectra for the calibration areas using
• use empirical line method to calculate gains and offsets for calibration to re- flectance with
• calibrate to reflectance using
to apply gains and offsets
• calculate histogram parameters using
• view spectra from the calibrated cube us- ing
and interactively select bad bands
• rotate image 180 ° to north using
• make BIP cube using
to allow efficient spectral viewing and analysis
Interactive Viewing and Analysis in SIPS_View
• display gray-scale image
• use Spectra Browse function to evaluate spectral character of images
• select polygons containing calibrated light and dark targets, load into View Spectra and compare to ground spectra used for calibration; validate calibration
• browse through several gray-scale images using the slider-bar to look for spectral differences
• produce a variety of color images based on absorption bands of known materials to locate areas with absorption features
• extract spectral slices to evaluate spectral changes along specific traverses
• use histograms and linear stretches to pro- duce enhanced images and save to color
• produce color printer quicklook copies for reference
• use Spectra Browse function to examine individual spectra and the Spectra Aver- age function to extract average spectra for areas showing color differences and save to spectra library
• examine spectra using View Spectra op- tion, load spectral libraries and compare to determine minerals and other materials
• use image endmembers in saved image spectral library to perform SAM analysis within SIPS View
• select endmembers from library
• edit spectral ranges
• view results using SAM viewer
• use sliders to highlight areas of high match
• save to SAM results cube
• exit SIPS View and reload SAM results cube as color image to show mixtures
• save results to files for filmwriter
• display single endmembers as density- sliced images and save to files for film- writer
• Reexamine endmembers based on results of SAM analysis using BSQ and BIP cubes in SIPS View
• Unmix image using unconstrained un- mixing
• select endmembers
• evaluate degeneracy of the library and adjust if required
• evaluate abundance images, error im- ages and sum
• revise endmembers if required
• Unmix image using constrained unmixing
• reselect endmembers if required
• evaluate degeneracy of the library and adjust if required
• evaluate abundance images, error im- ages and sum
• revise e n d m e m b e r images if required
• rerun unmixing if required
• import gray-scale images, color compos- ites, SAM analysis results images, unmix-
Spectral Imaging Processing System
ing results images into standard image processing software, or IDL for further image analysis, classification, statistics, etc.
• produce hardcopy output using filmwriter
• Register images to map base
• transfer results to GIS system for further analysis with results from other images, field mapping, field spectra, and labora- tory analytical work
SIPS is being released to organizations outside
CSES for analysis of imaging spectrometer data such as that produced by AVIRIS. To promote scientific use of these data, SIPS will be provided free of charge or royalties to any organization interested in use of imaging spectrometer data.
CSES plans to continue development of these programs and retains the title and copyright to the software, documentation, and supporting ma- terials. Recipients of this software are required to execute a memorandum of understanding (MOU) provided by CSES that specifies in detail all of the associated conditions. To get a copy of the software agreement contact Kathy Heidebrecht or Fred Kruse by electronic mail ([email protected] ado.edu), phone (303-492-1866), or FAX (303-
SIPS is an integrated software system for analysis of imaging speetrometer data. SIPS is designed to take advantage of the inherent strength of im- aging spectrometer data, simultaneous high reso- lution spectral measurements and spatial display.
It provides the basic capabilities to proceed from raw radiance data, through calibration, to inter- active viewing and analysis, to quantitative results and hardeopy output. It provides utilities for input of data, data formatting, data calibration, and other common image processing procedures.
Data visualization tools are provided for rapid, exploratory analysis and numerical tools are pro- vided for quantitative modeling.
SIPS makes possible routine display and anal- ysis of a volume and complexity of information that up until now have detailed analyses difficult.
It has been used for analysis of imaging spectrom- eter data from AVIRIS, GERIS, and GEOSCAN, and to look at other multispectral data sets from
Landsat MSS, Landsat TM, and SPOT. The proto- type interactive software system using IDL on
UNIX-based workstations simplifies analysis of imaging spectrometer data by allowing scientists to rapidly interact with entire datasets. It is our hope that these tools will be useful across multiple disciplines and allow quantitative analysis that will lead to new scientific discoveries using imaging spectrometer data.
CSES is continuing to develop SIPS as a gen- eral tool for analysis of imaging spectrometer data.
One of the main goals of this effort is to modu- larize the program to allow users to add custo- mized functions.
HARDWARE / SOFTWARE REQUIREMENTS
SIPS runs on Unix-based workstations under ei- ther Motif or Openlook window managers in 8-bit color mode. The platforms and the software ver- sions on which SIPS has been tested are shown in
Table 1. IDL version 2.4 or higher is required.
While SIPS should work on any platform that supports IDL (with widgets), these are the only platforms that have been tested to date. For more information concerning IDL, contact Research
Systems, Inc., 777 29th St., Boulder, CO 80303. v
SIPS was originally developed as a means fiJr viewing and analyzing A VIRIS data. Many of the ideas and techniques incorporated into SIPS are the result of nearly 10 years' experience with imaging spectrometer analysis by principals at CSES. The SPAM and ISIS software provided some impetus towards the types of analyses we wanted to perform, and we would like to acknowledge this influence. SIPS, however, was developed from scratch using the IDL programming language to satisfy specific analysis requirements not available in any existing software package. The basic interactive package
(SIPS_View) was developed under funding from NASA as part of the Innovative Research Program funded research proposal
"Artificial Intelligence for Geologic Mapping," NASA Grant
NAGW-1601 (Dr. F. A. Kruse, Principal Investigator). Addi- tional support for documentation of SIPS and development of unmixing routines included as part of SIPS were supported respectively by NASA Grant NAS5-30552 (Dr. A. F. H. Goetz,
Principal Investigator) and by a NASA Graduate Research
Fellowship (Dr. J. W. Boardman). The interactive SIPS View program, version 1. O, was written by A. B. Lefkoff with ~ersion
1.1 additions by A. T. Shapiro, P. J. Barloon, and K. B.
Heidebrecht. J. W. Boardman wrote the SIPS spectral unmixing
Kruse et al. routine and the Spectral Angle Mapper algorithm. Continuing development and support of SIPS as a HIRIS-team resource is funded by NASA Grant NAS5-30552 (Dr. Alexander F. H.
Goetz, Principal Investigator).
Boardman, J. W. (1989), Inversion of imaging spectrometry data using singular value decomposition, in
IGARSS "89, 12th Canadian Symposium on Remote Sens-
ing, Vol. 4, IGARSS, Canada, pp. 2069-2072. (Available from the Institute of Electrical & Electronics Engineers,
Piscataway, NJ 08854.)
Boardman, J. w. (1990), Inversion of high spectral resolution data, in Proceedings, SPIE, Bellingham, WA, Vol. 1298, pp. 222-233.
Boardman, J. W. (1991), Sedimentary facies analysis using imaging spectrometry: a geophysical inverse problem,
Ph.D. dissertation, University of Colorado, Boulder, 212 p. (unpublished).
Boardman, J. W. (1993a), Spectral angle mapping: a rapid measure of spectral similarity (in preparation).
Boardman, J. W. (1993b), Sedimentary facies analysis using imaging spectrometry; a geophysical inverse problem (in preparation).
Center for the Study of Earth from Space (CSES), (1992),
SIPS User's Guide, The Spectral Image Processing System
v. 1.1, Boulder, CO, 74 pp.
Clark, R. N., King, T. V. V., Klejwa, M., and Swayze, G. A.
(1990), High spectral resolution spectroscopy of minerals,
J. Geophys. Res. 95(B8):12,653-12,680.
Elvidge, C. D. (1988), Vegetation reflectance features in
AVIRIS data, in
Proceedings, International Symposium on
Remote Sensing of Environment, Sixth Thematic Confer-
ence, Remote Sensing for Exploration Geology, Houston,
TX 16-19 May, Environmental Research Institute of
Michigan, Ann Arbor, pp. 169-182.
Gao, B. C., and Goetz, A. F. H. (1990), Column atmospheric water vapor and vegetation liquid water retrievals from
Airborne Imaging Spectrometer data,
J. Geophys. Res.
Geophysical and Environmental Research (GER) (1988),
Single beam visible~InfraRed Intelligent Spectroradi-
ometer (SIRIS) User's Manual, GER, Millbrook, NY.
Goetz, A. F. H. (1981), Spectroscopic remote sensing for geological applications, in
Proceedings, Society of Photo-
Optical Instrumentation Engineers (SPIE), Bellingham,
WA, Vol. 268, pp. 17-21.
Goetz, A. F. H. (1984), High spectral resolution remote sensing of the land, in
Proceedings, Society of Photo-
Optical Instrumentation Engineers (SPIE), Bellingham,
WA, Vol. 475, pp. 56-68.
Goetz, A. F. H., and Boardman, J. W. (1989), Quantitative determination of imaging spectrometer specifications based on spectral mixing models, in
"89, 12th Canadian Symposium on Remote Sensing, Vol.
2, IGARSS, Canada, pp. 1036-1039. (Available from the
Institute of Electrical & Electronics Engineers, Piscata- way, NJ 08854.)
Goetz, A. F. H., and Calvin, W. M. (1987), Imaging spectrom- etry: Spectral resolution and analytical identification of spectral features, in
Proceedings, Society of Photo-Optical
Instrumentation Engineers (SPIE), Vol. 834, pp. 158-165.
Goetz, A. F. H., and Davis, C. O. (1991), High Resolution
Imaging Spectrometer (HIRIS): science and instrument,
Int. J. Imaging Syst. Technol. 3:131-143.
Goetz, A. F. H., Vane, G., Solomon, J. E., and Rock, B. N.
(1985), Imaging spectrometry for earth remote sensing,
Golub, G. H., and Van Loan, C. F. (1983),
tions, John Hopkins University Press, Baltimore, MD.
Green, R. O. (Ed.) (1990),
Proceedings of the Airborne Visi- ble/Infrared Imaging Spectrometer (A VIRIS) Workshop,
JPL Publication 90-54, Pasadena, CA, 280 pp.
Grove, C. I., Hook, S. J., and Paylor, E. D. (1992), Laboratory reflectance spectra of 160 minerals, 0.4 to 2.5 microme- ters, JPL Publication 92-2, Pasadena, CA.
Huntington, J. F., Green, A. A., and Craig, M. D. (1986),
Preliminary geological investigation of AIS data at Mary
Kathleen, Queensland, Australia, in
Proceedings, 2nd Air- borne Imaging Spectrometer (AIS) Data Analysis Work-
shop, 6-8 May, JPL Publication 86-35, Pasadena, CA, pp.
Jet Propulsion Laboratory (1991),
Planetary Data System
Data Preparation Workbook- Volume 1, Procedures, Ver- sion 2.0, Pasadena, CA.
Kruse, F. A. (1987), Extracting spectral information from imaging spectrometer data: a case history from the north- ern Grapevine Mountains, Nevada / California: in
ings, 31st Annual International Technical Symposium, So- ciety of Photo-Optical Instrumentation Engineers (SPIE),
Bellingham, WA, Vol. 834, pp. 119-128.
Kruse, F. A. (1988), Use of Airborne Imaging Spectrometer data to map minerals associated with hydrothermally al- tered rocks in the northern Grapevine Mountains, Nevada and California,
Remote Sens. Environ. 24(1):31-51.
Kruse, F. A., and Dietz, J. B. (1991), Integration of visible- through microwave-range multispectral image data sets for geologic mapping, in
Proceedings of the Cinqui~me
Colloque International, Mesures Physiques et Signatures en
T~l~d~tection, 14-18 January, Courchevel, France, Euro- pean Space Agency, Paris, France, ESA SP-319, Vol. 2, pp. 481-486.
Kruse, F. A., Raines, G. L., and Watson, K. (1985), Analytical techniques for extracting geologic information from multi- channel airborne spectroradiometer and airborne imaging spectrometer data, in
Proceedings, International Sympo- sium on Remote Sensing of Environment, Fourth Thematic
Conference, Remote Sensing for Exploration Geology, San
Francisco, 1-4 April, Environmental Research Institute of Michigan (ERIM), Ann Arbor, MI, pp. 309-324.
Spectral Imaging Processing System
Kruse, F. A., Kierein-Young, K. S., and Boardman, J. W.
(1990), Mineral mapping at Cuprite, Nevada with a 63- channel imaging spectrometer,
Photogramm. Eng. Remote
Kruse, F. A., Lefkoff, A. B., and Dietz, J. B. (1993), Expert system-based mineral mapping in northern Death Valley,
California / Nevada using the Airborne Visible / Infrared
Imaging Spectrometer (AVIRIS),
Remote Sens. Environ.
Marsh, S. E., and McKeon, J. B. (1983), Integrated analysis of high-resolution field and airborne spectroradiometer data for alteration mapping, Econ. Geol. 78(4):618-632.
Mazer, A. S., Martin, M., Lee, M., and Solomon, J. E. (1988),
Image processing software for imaging spectrometry data analysis,
Remote Sens. Environ. 24(1):201-210.
NASA (1987), HIRIS, High-Resolution Imaging Spectrome- ter: science opportunities for the 1990s, Earth Observing
System, Instrument panel report, V. IIc, National Aero- nautics and Space Administration, Washington, DC,
Open Software Foundation Inc. (1989),
Guide, Cambridge, MA, 56 pp.
Pieters, C. M. (1990),
Reflectance Experiment Laboratory
Description and User's Manual, RELAB, Brown Univer- sity, Providence, RI.
Porter, W. M., and Enmark, H. T. (1987), A system overview of the Airborne Visible/Infrared Imaging Spectrometer
Proceedings, 31st Annual International Techni-
cal Symposium, Society of Photo-Optical Instrumentation
Engineers (SPIE), Bellingham, WA, Vol. 834, pp. 22-31.
Press, W. H., Flannery, B. P., Teukolsky, S. A., and Vetter- ling, W. T. (1986), Numerical Recipes, Cambridge Univer- sity Press, New York.
Research Systems, Inc. (1991),
IDL ® User's Guide Version
2.2, Boulder, CO.
Roberts, D. A., Yamaguchi, Y., and Lyon, R. J. P. (1985),
Calibration of Airborne Imaging Spectrometer data to percent reflectance using field spectral measurements, in
Proceedings, Nineteenth International Symposium on
Remote Sensing of Environment, Environmental Research
Institute of Michigan (ERIM), Ann Arbor, MI, 21-25
Singer, R. B. (1981), Near-infrared spectral reflectance of mineral mixtures: systematic combinations of pyroxenes, olivine, and iron oxides, J. Geophys. Res. 86:7967-7982.
Smith, M. O , and Adams, J. B. (1985), Interpretation of AIS images of Cuprite, Nevada using constraints of spectral mixtures, in Proceedings, AIS Workshop 8-10 April, JPL
Publication 85-41, Pasadena, CA, pp. 62-67.
User's Guide to Microlmage Software, Ver-
sion 4.0, Terra-Mar Resource Information Systems, Inc.,
Mountain View, CA.
Torson, J. M. (1989), Interactive image cube visualization and analysis, in
Proceedings, Chapel Hill Workshop on
Volume Visualization, 18-19 May, University of North
Carolina, Chapel Hill.
Vane, G. (Ed.) (1987), Imaging spectroscopy II, in
ings, 31st Annual International Technical Symposium, So- ciety of Photo-Optical Instrumentation Engineers (SPIE),
Bellingham, WA, Vol. 834, 232 pp.
Vane, G. (Ed.) (1988),
Proceedings of the Airborne Visible/
Infrared Imaging Spectrometer (AVIRIS) Performance
Evaluation Workshop, JPL Publication 88-38, Pasadena,
CA, 235 pp.
Vane, G., and Goetz, A. F. H. (Eds.) (1985),
Proceedings of the Airborne Imaging Spectrometer (AIS) Data Analysis
Workshop, 8-10 April, JPL Publication 85-41, Pasadena,
CA, 173 pp.
Vane, G., and Goetz, A. F. H. (Eds.) (1986),
Proceedings of the
2nd Airborne Imaging Spectrometer (AIS) Data Analysis
Workshop, 6-8 May, JPL Publication 86-35, Pasadena,
CA, 212 pp.
Vane, G., Green, R. O., Chrien, T. G., Enmark, H. T.,
Hansen, E. G., and Porter, W. M. (1993), The airborne visible/infrared imaging spectrometer (AVIRIS),
Sens. Environ. 44:127-143.
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project