Hi Globillumers --

Well, here are the results for the reference image experiment.
The bad news is that there were not a large number of
participants. The good news is that we got remarkable agreement
on the resulting image using four distinctively different
approaches. 

The four entries were from --

    Raphael Compagnon (COMPAGNON@eldp.epfl.ch)
     using SUPERLITE to calculate radiances, and "Radiance"
     to find the visible surface at each pixel

    Chris Christou (cgc@physiology.oxford.ac.uk)
      using a radiosity solution to get a radiosity
      for each surface, ray-tracing to find surface visible
      through each pixel and then gathering the radiosity
      result to calculate the pixel value

    me
      using my own Monte Carlo path tracing program

    Greg Ward (greg@hobbes.lbl.gov)
      using "Radiance"


Let R = (max value - min value) reported for each pixel,

R = 0 for  607 pixels
  = 1 for 1425 pixels
  = 2 for  341 pixels
  = 3 for   79 pixels
  = 4 for   20 pixels
  = 5 for    8 pixels
  = 6 for    4 pixels
  = 7 for    4 pixels
  = 8 for    7 pixels
  = 9 for    2 pixels
  = 12 for   2 pixels
  = 14 for   1 pixel

Under pub/holly on tiber.nist.gov, I have posted an image
fin_res  in a number of different formats. The red channel
in fin_res is the minimum value reported at each pixel, 
the green is the average value, ane the blue is the maximum
value. 

Since tiber isn't meant to be an archive -- it's supposed to
be used for transferring data via anonymous ftp to people
outside of NIST -- I can only plan on leaving the results
there for about a month.

-- Holly

 

   