GEOG 751: Digital Techniques of Remote Sensing

Exercise #9

Spectral Enhancement:
Image Indices and Principal Components Analysis

Objectives

Image

Part I Image Indices

ERDAS Imagine offers several well known transformation indices used for spectral enhancement. Vegetation indices are used to measure the presence and condition of green vegetation. These indices are based on differences in the response of vegetation as measured in the NIR and red regions of the spectrum.

For this part of the exercise, open the Image Interpreter menu and select Spectral Enhancement. In the Spectral Enhancement menu, select the Indices... option. This should open the Indices dialog box which allows you to specify the sensor and has a variety of indice functions. Select charleston_tm_1994-02-03.img as the Input File and give the Output File any name you choose (i.e. charleston_tm_1994-02-03_ndvi.img). The Coordinate Type should be set to map and Sensor set to Landsat TM. Set Select Function to NDVI and turn on the Stretch to Unsigned 8-bit (this saves space). Leave the other variables in their default state. Note the function being used and the bands that are incorporated into this function. Select OK after you have set all the variables. View the results and answer the following question. (Note: you might want to display an infrared color composite of charleston_tm_1994-02-03.img for comparison.)


Part II Principle Components Analysis

Principal Components Analysis (PCA) can be used by image analysts as a data reduction technique whereby the information content from a number of bands is compressed into a few principal components. In other words, PCA can be used to reduce the dimensionality of the data while minimizing loss of information. In addition, PCA images may be more easily interpreted than the conventional color infrared composite. For your reference, during the first pass, a covariance matrix of the input bands is computed. This covariance matrix is then used during the second pass to compute the principal components or eigenvectors.

Principal Component Analysis procedures using Imagine:

Open the charleston_tm_1994-02-03.img file (4,3,2) in an imagine viewer. Next, under the Interpreter menu button select Spectral Enhancement then Principal Comp... When the Principle Components dialog box appears, use charleston_tm_1994-02-03.img as the input file and give an appropriate name for the output file (i.e. charleston_tm_1994-02-03_pca.img). Select File as the coordinate type. In this part of the exercise, we will be processing the entire scene. Note that it would also be possible to subset the image using an inquire box or coordinates that the user types in. Leave the Data Type unchanged, i.e. Input: Unsigned 8 Bit and Output: Float Single. Leave all Output Options in their default state. In both the Eigen Matrix and Eigenvalues sections select the Write to File: option. A default name for each should appear in their respective screens. Finally, select 7 for the Number of Components Desired:. If you are interested in viewing a graphical representation of the PCA process (it might help you understand it more, then again it might not) click on View before you complete the enhancement by clicking on OK.

When the processing stops, click OK and then open a new viewer. In that viewer open your freshly created PCA image using File - Open - Multi-Layer. Seven windows should now open, showing you the individual six PCA "bands" and one composite "band". You can also open another viewer and display color composites of different combinations of the components containing most of the variance to provide more information. Do so by opening the PCA image not as a multi-layer but as you would a normal composite image. Choose the RGB combination that you feel most exemplifies the majority of the information present given your knowledge about information in each of the six layers from observation of the individual layers.

To answer the next few questions, you will need to study the images and the files containing the Eigenvector matrix (*.mtx) and Eigenvalues (*.tbl). To view, and understand, the two output files, open each of them in a text editor, such as Notepad or Wordpad. The *.mtx files when opened will more than likely appear as a set of seemingly random numbers. To make sense of them, enlarge the screen so that the numbers are aligned in seven columns, one for each of the principal components (you may have some negative numbers). If you have successfully completed this task you should have a table of seven colums with seven rows each. The columns, as said, correspond to the seven principal components and the seven rows correspond to the seven bands of TM data. The numbers represent a factor score (Eigenvector) that each band contributed to the individual component. If band 4 contributed close to 1.0 to the component, one could argue that this specific component is good for measuring vegetation cover. The *.tbl file gives you the Eigenvalues for each of the six principal components. The total of these figures will give you the total variance.

To complete this part of the exercise, create another map composition and display the three chosen components (in question 7) as monochrome images along with a color composite of these same three components. Be sure your map composition has the appropriate annotation.






Back to Exercises

Last Modified: September 27, 2004