[argyllcms] Re: Optimizing contrast to fit source gamut into device

  • From: Nikolay Pokhilchenko <nikolay_po@xxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Wed, 16 Dec 2009 10:31:01 +0300

Graeme Gill wrote

> Nikolay Pokhilchenko wrote:
> > Hello, Graeme! For a long time as I start using the collink in gamut 
> > mapping mode, I
> > want ask You a contrast optimizing feature - some kind of a gamut-depended 
> > "auto
> > levels". In some cases the source gamut is exceeds the destination gamut 
> > only from top
> > For example take a look at
> > http://www.colour.org/tc8-03/images/ski_pg/Ski_TC8-03_abs.tif test image. 
> > While
> > performing of conventional gamut mapping by collink, the output image is 
> > much more
> > desaturated as it may be with contrast adjusting.
> Hmm. I'm not sure this is really a good fit for gamut mapping. It seems to me
> that this is more like "exposure adjustment", and therefore falls outside
> what "collink" is intended to do.

I'm not completely agree here. Since gamut of source image depends of exposure 
value and the photographer can give a freedom to mapping algorithm to change an 
exposure to better fit. In general, the given camera exposure mainly selected 
in an effort of the avoidance of camera sensor saturation, not for printer 
gamut effort. From RGB-camera point of view there may be no problem to capture 
without saturation a bright color-saturated object. But there will be problem 
to map such image to subtractive, passive print. If the darkest area of image 
is not at the lower boundary of device gamut, the whole image can be shifted 
down (preferably in Jab space) in an effort of bright and saturated colors. I 
think as soon as we do mapping in true perceptual space, the lights and the 
darks have an equal rights. If the lights are highly compressed by device gamut 
and darks are not, it _MAY_ be useful to shift the whole input space down in 
Jab. A clarifying: the whole scale shift didn't mean th
 e contrast changing. On the contrary, it helps to preserve the contrast for 
saturated image areas.

> Argyll gamut mapping very deliberately
> bases it's neutral axis and white & black point adjustments on the source
> colorspace, on the assumption that any image within that colorspace has
> been adjusted appropriately with regard to these three attributes. Adjusting
> these three attributes + general contrast is something that would need to be 
> more
> heuristic, and I think would be better in a separate "auto image adjust" type
> utility.

I'm not completely agree here too. There should be the key for Argyll gamut 
mapping which gives or not gives the mapping algorithm the freedom to tune 
input gamut position along an L axis.

Thank for Your attention, Graeme!

Best regards,

Other related posts: