[argyllcms] scanner profiles and white points

  • From: Tim Gray <tgray@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Thu, 27 Jan 2011 14:14:52 -0500

Hi, new user of Argyll CMS. I'm trying to profile my scanner (Nikon Coolscan V) using IT8 slide targets. I'm having some issues with blown highlights and wanted to know if I'm doing things right.


First I obtain a linear raw scan of my target (using Vuescan). I save this as a tif. I then create the ti3 file with the following command:

        scanin -v -G 1 kodachrome.tif ref/it8.cht E090805.txt

After that I call colprof and run it on the ti3 file:

        colprof -v -D"NSV - Kodachrome" -qh -bn -ax kodachrome

This generates an icc file.  All good so far (I think).

I then take a linear scan of a slide and bring it into Photoshop and assign it my new profile. At this point I get blown out highlights on some slides.

I think this stems from the fact that the white on my target scan is not as bright as the white on some of my slides. I'm not sure why this would be - I would assume that the white to the left of GS1 (GS0?) is pretty much the dmin of the target slide material.

I've figured out several ways to work around this, but they all seem like kludges; I'd like to figure out a proper way to deal with this. One thing I've seen that preserves highlights is to use a 'absolute colorimetric' conversion to my target space (AdobeRGB for what it's worth). However, the appearance of many other colors changes. The other thing which seems to preserve highlights is to run the levels command after assigning the scanner profile but before conversion to my editing space. In the levels dialogue I change the 'output levels' white point until the highlights come in. After converting to the target space, everything stays put.

Lastly, I noticed that the highlights are reigned in some if I generated an icc with the command:

        colprof -v -D"NSV - Kodachrome" -qh -bn -ax -U2.0 kodachrome

This appears to work, but it's kind of a guessing game as to which scaling factor to use. Just for reference, I also did this whole procedure using non-linear scans and the same issues occur.

Any ideas would be appreciated. Should I be modifying my original target scan in some manner? Something else?

Thanks,
Tim

Other related posts: