[argyllcms] newbie questions on profiling input devices
- From: ks <transients@xxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Mon, 28 Nov 2005 08:11:18 +0100
I'm trying to create an icc profile for my camera. The profile is going to
be used with a raw conversion software. The software has a mode to by
bypass all colormanagment and output unmodified rgb tifs. I have made
shots of an it8 target (wolf faust), converted it to tif in this 'bypass
cm mode' and process it with argyll. The input profile created is later on
used directly by the raw converter.
I already have have created profiles using the scanin and profile commands:
scanin -v %1.tif %it8patch% %it8targ%
profile -v -D%descrip% -C%copyright% -M%model% -qu %1
While overall this works very well, there are some flaws in the profile
and I don't know if and how the argyll tools can help me to improve it. SO
here are my questions related to this:
1) The range of reflectivities in the it8 target does not span the full
dynamic range of the camera. I assume that the icc profile must provide a
transformation for all incoming rgb values. It follows that profile.exe
must partly extrapolate some points in the CLUT table. How should I deal
with this? What is best practice when using profile, leave the it8.tif as
is, use a level tool (PS) to shift the histogram to the max (without
clipping), or even stretch the whole histogram to fill the whole range?
From my experiments it looks that the latter procedure gives worse
2) Are there other command line switches in profile that are worth giving
a try? I tried the -t/-T options but nothing(?) seems to change. Do I have
to use it like '-t2' or '-t 2' ?
3) The profiles build into the raw converter exactly preserve gray values
at all levels. In other words the transformation for a triple with r=g=b
is always r'=a*r, g'=a*g and b'=a*b. This is important for white balancing
and to prevent color casts. Is there a way to apply such a constraint to
the profile generation or any other trick? I know in an ideal world this
should be the outcome of the regular unconstrained profiling but I guess
that it is very hard to achieve in practice. Didnt I tried hard enough?
4) -p,s,S options: They look potentially useful for my task but have yet
no idea what to do with them. Can I use the -p option to apply an
additional 'curve' or 'level' (in PS terms) transformation. Is there a
tool to easily create such abstract profiles?
5) (related to 1) It might happen that some colors outside the range of my
it8 ref target are not transformed very well. Is there a way to tweak the
exisiting profile to make improvements in these 'extreme' color ranges.
I'm sorry to have bothered you with my probably very basic questions. Any
help is highly appreciated!
Other related posts:
- » [argyllcms] newbie questions on profiling input devices