Level 2 Help for EIGENVEC

Input datasets containing registered images, or a single input dataset 
in MSS format.

Output datasets containing transformed images, or a single output
dataset in MSS format.

The standard Vicar size field ( starting_line, starting_sample, 
number_of_lines, number_of_samples).

Statistics are gathered using only every n'th line of the image. If
the auto-scaling option of XFORM is invoked, every n'th line is used to
determine the scaling parameters.

Statistics are gathered using only every n'th line and n'th sample of the
image. If different line and sample increments are desired, use INC to
specify the sample increment and use LINC to override the INC line increment
value.

The correlation matrix is used for computing eigenvectors instead of the
covariance matrix.

When the DSTRETCH keyword is used, the output matrix will be a composite
matrix that describes the color decorrelation stretch transformation. This
matrix is the product of the eigenvector matrix, a variance equalization
vector (which has the effect of contrast stretching), and the transpose of
the original matrix (for the "back rotation" step).

This parameter has effect only when used with the DSTRETCH parameter. The
DSTRETCH operation consists of a rotation into principal component space,
rescaling to cause variance equalization among components, and the reverse
rotation out of principal component space. DSCALE applies an additional
scaling at the variance equalization stage. This can be used to suppress
components known to be noisy, at the expense of re-introducing some 
correlation among bands.

A pixel is excluded from use in determining the transformation if the 
DN is equal to 'EXCLUDE' in each channel. If the auto-scaling option of 
XFORM is used, these pixels will also be excluded from the computation of
scaling parameters.

Denotes the input data is in MSS format and contains 'MSS' number of 
channels of data.

If the input is in MSS format, use only the listed channels. The default
is that all channels are used.

Denotes that that output is to be in MSS format, and will contain 'MSSO'
components (bands) of data.

The output will consist of the specified components. The default is to
output the first n components, where n is the number of output datasets,
or for MSS format output, the value of MSSO.

The eigenvector transformation matrix will be saved as a VICAR2 parameter
dataset with the name denoted by the SAVE parameter.

The transpose of the eigenvector transformation matrix will be saved as a
VICAR2 parameter dataset with the name denoted by the TSAVE parameter. This
matrix produces the rotation necessary to return to the original coordinate
system.

Sets of (Starting_line, Starting_sample, Number_of_lines, Number_of_samples)
are given to define subareas used to generate the image statistics. If
auto-scaling is used in XFORM, only the first subarea listed will be used
to gather the auto-scaling statistics. The default is that the entire image 
is sampled.

Integer specifying the dimensionality of the data.  This parameter is 
only used when multi-variate data is input through the parameter field
(DATA).

1 to 500 data values may be input through the parameter field. That is, the
number of case times NVAR must be no greater than 500. No input data sets 
are used in this mode.  The covariance matrix and the eigenvectors for this 
set of data are computed and printed.