Test results:

The current algorithm was used to integrate a set of peaks from TOPAZ runs
5637-5644.  The results were run through GSAS restricting peaks to minimum 
integrated intensities of 70,60,45, and 30 and peak centers within 24, 20, 15,
10, and 5 pixels of the edge.  The pixels within 5 pixels of the edge had no
data, so the "edge" was at 5 pixels.

   First, the GSAS result(R(F**2) using peaks with min integrated intensities of 
70 and within 24 pixels of the edge, was .092, which agreed with the results from 
other algorithms for integrating peaks. Indeed it was a little lower.  Also, this 
R factor is quite high.  TOPAZ is working on basic detector problems.  The panels
are not "balanced". When balanced for GSAS, the R factor was approximately 2% less. 
Other adjustments could also decrease this R factor.

    The GSAS R value for peaks with min integrated intensities of 70,60,45, and 30
were the same( up to .001).  Using peaks at least 20  and 15 pixels from the edge 
yielded a GSAS R value of .091.  For 10 pixels the R factor was .097. Unfortunately,
using peaks whose centers were 5 pixels or more from the edge gave a GSAS R factor 
of .287.

    Conclusion: The algorithm works with weak peaks.  Using all peaks within 5 pixels 
of the true "edge" of a panel is "good enough".  These were the results from a first 
pass at improving the results of peaks that a really close to the edge.  More work 
would be necessary to improve these values, if possible.  Unfortunately, time and
money has run out for further improvements at this time.  Also, the instrument scientists
are working on more basic problems, there may be fundamental detector problems with 
edge values, and other algorithms may be used for these edge peaks.