Re: [UFRaw-Devel] [UFRaw-CVS] ufraw ufraw.h, 1.40, 1.41 ufraw_conf.c, 1.46, 1.47 ufraw_developer.c, 1.27, 1.28 ufraw_ufraw.c, 1.76, 1.77

[Greg T. <gd...@ir...>]

  Modified Files:
          ufraw.h ufraw_conf.c ufraw_developer.c ufraw_ufraw.c 
  Log Message:
  Normalize the exposure for Canon DSLRs.

Interesting - I was just about to write about a perhaps similar issue
that I noticed with the Nikon D200.

I took pictures in RAW (which has embedded JPG BASIC) of gray/white
cards (Kodak R-27: 18% and 90%), spot metering on gray.  I also
measured luminace in cd/m^2 with a spot meter, and was trying to
understand the relationship from lumiance to chosen exposure.  As a
side effect, I compared pixel values of the JPG (with all image
settings set to standard, not AUTO) and the jpg from ufraw with
everything set to default.  They were way off, with the JPG being much
brighter.  I also found that the ratio of gray to white was compressed
in the JPG compared to photometrically accurate, and closer in RAW.


So, this raises three perhaps separate issues:

1) Should we declare that the right behavior for raw conversion with
all default (0) paramaters is to produce an image that more or less
matches the camera when contrast/etc. is set to 'normal' (vs auto)?
If so, should there be a different gain setting vs what I think is the
assumption that 1023 corresponds to 255.0 or something like that?

2) It seems the midtone values are off at least half a stop compared
   to the camera's behavior.


3) scene luminance ratios are greater than display and print luminance
ratios, and there's a long tradition in photography of mapping them
perceptually; this is what the zone system is all about.  The D200 is
clearly doing this at normal, compressing 18/90 to more like 28/90,
which leaves more room for shadows/highlights without clipping.

So perhaps there should be a contrast control for this, much like the
'normal', 'normal plus' nad 'normal minus' that Adams discusses in
_The Negative_.


Finally, I'd like to be able to see luminance in cd/m^2 in the spot
value as well, but until the above is better understood I'm afraid
that will be a bit sloppy.

Udi: can you explain what the canon adjustment you just put in is?  Is
it the same as my point 2 above?

-- 
    Greg Troxel <gd...@ir...>