Program by John Pierce, Jonathan Greenberg and George Scheer
Property of the CSTARS Lab at the University of California, Davis. Please ask for permission before distributing this script.

Introduction:

MATLAB function for performing continuum removal analyses on hyperspectral datasets using the trapezoidal method. This is calculated by taking the ratio of the area under the absorption curve and the area of the trapezoid formed by the two shoulders of the absorption feature. Here, the value is returned as a percentage (0 <= absorption value <= 100).

Note (3/15/02): If you may want to consider using 1 - continuum removal output to generate a postive slope when comparing the CR output to an amount of material to save yourself some effort explaining that the lower the ratio, the higher the absorption.

System Requirements:

This program needs quite a bit of memory, but you can modify this somewhat by lowering the "maxPixelRead" value.

• Unix users: make sure you have enough swap space (talk to your sysadmin).
  
• Mac users: you'll want to turn virtual memory WAY up and then allocate more preferred memory for the application. To do this, SINGLE click on the MATLAB application and hit Command-I ("Get info" if you want to use the File menu). Older OSs will see the "Memory requirements" dialog-- change the "Preferred Size" to a much larger number, depending on your input filesize. Close the MATLAB info window and boot the program. Newer OSs should click where it says "General Information" and change it to "Memory". Follow the above steps for upping the memory.
  
• Windows users: make sure you have enough virtual memory.

With enough memory, this can be fairly speedy, but if it starts using virtual memory it can take awhile. Be patient. Go get some coffee and come back later.

Instructions:

1. Unzip the file with your favorite decompressor-- Unix users use "unzip [filename]", Macintosh users can use Stuffit Expander and Windows users have a myriad of decompressors (usually WinZip is the unzipper of choice).

2. Place the continuumRemoval.m file in your MATLAB:Toolbox:local folder.

3. You will need to first create a list of the wavelengths in BAND order of the sensor. A down-and-dirty way of doing this is to open the header of the image file and copy and paste the wavelengths into a separate text file. Each wavelength should be on a separate line. We used a text editor to search and replace all the commas with returns. Remember that it is important that you do NOT sort the wavelengths-- sensors like AVIRIS record data somewhat out of order, but we account for this in the program.

Note: we have included the text file for the full AVIRIS spectrum with this release, named "aviriswave.txt".

4. The command to initiate the function is "feval('continuumRemoval',inBipFile,inWaveFile,samples,lines,bands,header,minWave,maxWave,maxPixelRead,outBipFile)".

• inBipFile: the filename of the image to analyze in BIP format.
  
• inWaveFile: the unsorted list of wavelengths from the sensor, we have included "aviriswave.txt" for those using AVIRIS images that have not been subsetted.
  
• samples: the number of samples in the image (often the horizontal width in pixels-- check the header of the image if you aren't certain what this number is).
  
• lines: the number of lines in the image (the vertical width in pixels)-- again, check the header.
  
• bands: the number of spectral bands in your image. Check your header.
  
• header: the header offset-- usually just 0. Once again... check the header.
  
• minWave and maxWave: the local maximum surrounding the area the analysis will be performed on. This should be determined through literature reviews or analysis of your spectrum prior to running this analysis.
  
• maxPixelRead: this was implemented as a memory management function; try a value around 10000. This is the maximum number of pixels read into memory at a time.
  
• outBipFile: the continuum removal output-- it will be the same number of samples and lines in BIP format, band = 1, header offset = 0 and the data type is double precision.

Change Log:

1.1: Added "maxpixelread" argument which reduces memory usage significantly.
1.0: Original release

Bugs:

• Many images are given in different numerical formats, so it may be neccessary to modify the 'ieee-le.l64' part at line 77:
  fid = fopen(inBipFile,'r','ieee-le.l64');
  At some point, I will probably implement a better routine for determining this value. As it stands, I would recommend starting with this setting, and then working through the other possible numerical settings until the output looks correct. The output will be unrecognizable, more or less, if this setting is wrong so it makes finding the correct one a bit easier.
• In ENVI, try using Host(Intel) as the Byte Order when importing the output.

Contact:

Jonathan Greenberg, greenberg@ucdavis.edu.

The software on this page has been created by CSTARS members, and may be freely downloaded by outside users. However, if you wish to distribute or use the source code of these applications, please ask permission from the author(s) first.