[argyllcms] icclink -G and source gamuts /-profiles

  • From: Klaus Karcher <lists@xxxxxxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Sat, 10 May 2008 16:34:24 +0200

Hi,

In the context of my experiments with tiffgamut/icclink -G, I stumbled upon something I can't understand:

As Photoshop's "Reduce Noise" filter doesn't work with "one way profiles" (like many camera or scanner profiles), I had to convert my source image to another colorspace. To avoid clipping and quantisazion errors, I decided to use Adobes ProPhoto RGB and 16bpc data. I was very surprised to see differences between the results.

To verify this differences and preclude errors, I made a second test:

I made a matrix profile based on ProPhoto RGB with the same primaries and whitepoint, but gamma 1 instead of 1.8.

I converted my source image to

a) ProPhoto RGB
b) my Gamma1-ProPhoto RGB

in Photoshop, relative colorimetric, without BPC in both cases.

I created gamut files with tiffgamut -ij from (a) and (b) and verified that the image gamuts are basically the same.

then I built two DLPs:

icclink -qh -ip -cmt -dpp -G [a|b].gam [a|b].icc [target].icc [a|b]_link.icc

and used them with cctiff:

cctiff [a|b]_link.icc [a|b]_source.tif [a|b]_converted.tif

I expected images a_converted.tif and b_converted.tif to be nearly identical, but they are very different (see screenshot).

What's going on?

TIA,
Klaus

JPEG image

Other related posts: