RE: Bad pixel determination with better software

[<aah@nofs.navy.mil>]

There are many families of bad pixels, depending on
the device.  For the IR world, we do the following
experiments to look for bad pixels.
  (1) dark frames at different exposure times, keeping
      the temperature constant.  When you histogram
      the response, there will be some hot pixels that
      stand out from the crowd.
  (2) dark frames at different temperatures.  Some pixels
      are very sensitive to temperature.
  (3) linearity set.  Use a constant illumination source
      and take exposures from low level to saturation.
      Histogramming the responses will show pixel families
      that are nonlinear or always hot/cold.
  (4) noise sets.  Take many exposures at a constant illumination
      level.  Some pixels will exhibit higher noise.
There are others, but this gives you a start.  The hard part
is determining the dividing line between 'good' pixels and
'bad' pixels.  For the Loral chips that I have seen from
Tom, there really aren't many bad pixels and it probably is
not worth a lot of effort in classifying them beyond the
visual inspection level.
Arne