Help for CALFIT

PURPOSE:

EXECUTION:

Example

calfit INP=(C1,C2,C3,C4) OUT=(A,B) CALVAL=(0.0,5.0,10.0,20.0)

OPERATION:

For each pixel, a linear calibration is calculated by least squares fit
on the calibration images.  Then, the average value for each pixel is
calculated for the first and last calval values.  Then for each pixel,
a linear transformation is calculated for compounding the lsq linear to
the average values.  This compounded linear transformation is output
to the output images (y=ax+b) where a values are real values in the
first output image and b values are real values in the second output
image.  A new image from the sensor can then be calibrated with these
two real images (A and B) by applying the VICAR routine f2 and the 
function IN1*IN2+IN3; where IN1 is the A image, IN3 is the B image and
IN2 is the sensor image to be calibated.

The calibration can be dropped and the least squares fit for each pixel
can be output in the A and B images.  Select mode=fitonly for this option.
Also, if numerous extreme values (dead sensor pixels) are ruining the
average value, the median value can be selected.

(alz comment) I'm not sure, but dead pixels may have been causing lsq fit
error returns.  Adespike of the inputs cured but too expensive.  So I
intercepted the lsq error message and output (0,0) as gain and offset.

Quadratic fit is performed if a third output file name is given. Can be
used only for the FITONLY mode.

The INVERSE keyword causes the program to invert y=ax+b solution
into x=ay+b solution.  Can be used to invert quadratic as well.
TIMING: 

TBD

 ORIGINAL PROGRAMMER:    A. Zobrist          12 Feb 2006
 COGNIZANT PROGRAMMER:   A. Zobrist          12 Feb 2006
 
 REVISION HISTORY
  2006-02-12 AZ  Initial version
  2006-12-18 AZ  Quadratic fit added
  2006-12-19 AZ  Inverse keyword added
  2008-01-03 WLB switched to USES_ANSI_C AND LIB_CARTO; misc cleanup