Dave Wagner wrote:
I finally resolved my problem. I was using the sRGB profile from http://www.color.org/srgbprofiles.xalter with BPC for my working profile in Gimp. When I used the sRGB profile without BPC, the shadow detail returned. So, here's a final clarification question: Which combination is "correct," assuming that gimp is not doing BPC... 1) Use sRGB_no_BCP as a source profile to argyll's profile utility AND sRGB_no_BCP as the RGB working space? 2) Use sRGB_with_BPC for profile generation, sRGB_no_BCP for working space? 3) Use sRGB_no_BCP for profile generation, sRGB_with_BPC for the working space?
I'd use either the sRGB_IEC61966-2-1_noBPC.icc or a real sRGB profile, such as the one from HP that was at www.srgb.com before the site disapeared (Size 3144 bytes, Date/Time 9 Feb 1998, 6:49:00) consistently. Don't mix and match unless you know exactly what you're doing. The sRGB_IEC61966-2-1_withBPC.icc represents the actual sRGB colorspace, although it has a faulty black point tag (the black point tag doesn't match the response in the table). Because it's faulty, don't use it. The sRGB_IEC61966-2-1_noBPC.icc doesn't represent sRGB - it has a non-zero black point, although the black point tag is correct. So this is a consistent profile which is why it works, although it doesn't represent the correct definition of the sRGB colorspace. Frédéric Mantegazza wrote: > There is something I don't understand: how a profile alone can do BPC? I > thought this could only be done by the CMM, using source and targer > profiles... It appears that the sRGB_IEC61966-2-1_withBPC.icc profile is a hack. It has a black point tag that is inconsistent with the actual shaper/matrix behaviour. I'd imagine with a particular CMM when linked with a particular output profile, this may have some sort of desirable behaviour, but it is fairly useless in practice because it's faulty. Matrix profiles can have only one intent, and it should be the base colorimetric intent, because that can then be used to create other intents. Trying to twist it into a partial perceptual intent is buying trouble. You can do "black point compensation" or rather gamut mapping with a Lut based profile (which is what Argyll profile -S is doing), and the perceptual intent has it's own table, leaving the colorimetric intent alone. Graeme Gill.