[argyllcms] Re: False colors with shaper/matrix display profiles
- From: Oliver Schulz <whisp@xxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Wed, 30 Aug 2006 17:49:20 +0200
> > When i generate shaper/matrix profiles, however, and then
> > use these profiles (with Scribus, or with tifficc to manually
> > correct images for the displays), i get bright
> > false colors in some very dark image areas. Using
> > -qh with 'profile' reduces these areas, but the problem
> > itself persists.
>
> Does this occur if you use (Argyll's) cctiff ?
I just checked, no, it does not.
# cctiff -e1 -ir sRGB.icm -ir DisplayA.icm input.tif output.tif
works fine.
The results of doing
# convert -profile -profile DisplayA.icm input.tif output.tif
and
# tifficc -i sRGB.icm -o DisplayA.icm input.tif output.tif
look exactly the same in most areas, but have the bright false
colors in some dark areas.
> Without the profiles you are using, I'm only guessing,
Oh, yeah, forgot that info in my original posting:
Input profile was the standard sRGB profile in all cases.
> but it sounds rather like a bug in cctiff that I fixed
> in V0.52, that was triggered when a profile created device
> values values > 1.0 for certain PCS inputs, causing pixel
> value wraparound. The profile triggering it is probably
> actually the other profile you are using, a gamma/shaper
> profile, since the gamma curve doesn't naturally
> clip the device values the way the shaper curve
> lookup does.
> I'd need to look at both profiles to be sure though.
>
> If in fact it is the same type of thing, then
> changing Argyll profile won't fix this, since it's the other
> profile that triggers the problem, and in fact it's
> a problem not with the profile, but with the pixel
> conversion engine Scribus and tifficc are using (lcms ?).
Yes, Scribus, tifficc and ImageMagick all use lcms, i think.
So, a bug in lcms?
Oliver
Other related posts:
