On 08/30/2015 11:37 AM, marti.ma...@littlecms.com wrote: > > Quoting Elle Stone <ellest...@ninedegreesbelow.com>: > >> Standard RGB working spaces are defined by Red, Green, and Blue >> primaries, a color space white point and a TRC that is the same in all >> three channels, plus chromatic adaptation from the color space white >> point to the profile illuminant. > > Maybe this is what standard working space means to you, buy there is > nothing in the ICC spec that refers to "workspace" profiles.
I started refering to "standard RGB working space" followed by a brief description of the characteristics of commonly used RGB matrix working spaces so as to not confuse a person connected with the ICC. That person took me to task for misrepresenting the new color.org LUT versions of the sRGB color space, in an article I wrote on the many variants of the sRGB *matrix* color space that float around open source and proprietary editing software: Will the Real sRGB Profile Please Stand Up? http://ninedegreesbelow.com/photography/srgb-profile-comparison.html I'm pretty sure that if you asked a person who uses programs like Krita, GIMP, and PhotoShop to name some standard RGB working spaces are, they'd say something like "sRGB, AdobeRGB, ProPhotoRGB". And they wouldn't mean the new color.org LUT versions of the old "legacy" simplified display matrix profiles, though many people download those new color.org profiles without understanding what they are getting. > What many > image editing apps are actually using are oversimplified *display* > profiles. Well, an awful lot of people manage to edit their images in various "oversimplified display profile" color spaces. I never suggested that sRGB makes a good display profile. It makes a fairly awful display profile unless you have a high end display that can be calibrated to match the sRGB color space. And even then surely there are residual issues with display black points, which surely vary from one display to another even with high end displays. > Whilst this may work on some situations, the profiles were never > intended for this purpose. Well, oftentimes something is invented and then used for purposes the inventor didn't have in mind. See Andrew Rodney's 2006 paper "The role of working spaces in Adobe applications" (https://www.adobe.com/digitalimag/pdfs/phscs2ip_colspace.pdf), in which he went to great pains to differentiate between device profiles (which typically are not suited for image editing) and working space profiles, which fall into the ICC's category of display profiles. The simplification is exactly what makes the "simplified device profile" suitable for use as an RGB working space when image editing. Is there is a currently viable alternative for using simplified display profiles like AdobeRGB1998, Rec.2020, ProPhotoRGB, and their linear gamma counterparts, for image editing using applications like Krita, GIMP, and PhotoShop? As "simplified display RGB working space profiles" are so widely used for image editing, maybe the ICC could give them a name all their own, to differentiate them from true display profiles? > > Just to enumerate some issues the matrix-shaper sRGB has: > > - It has no intents Well, "display profiles" (whether used for devices or as working spaces) used to have relative and absolute, didn't they? Until the V4 specs came out and the white point tag got relegated to always being D50. Right now the only effective intent for converting to a "simplified display matrix profile" seems to be relative colorimetric. > - If we assume it works in relative colorimetric, then black scaling is > wrong Well, as a working space, you kinda want a zero black point. As a display device profile, that's an entirely different matter. > - If we assume it works in perceptual, then the gamut mapping is very > poor (just clipping). Why would anyone assume a matrix profile worked in perceptual intent? Is this when converting to or from sRGB? When converting to the sRGB matrix profile, if you ask for perceptual intent, you get relative colorimetric, which clips. All conversions to the standard "simplified device" RGB working spaces clip. > - The black point is wrong in any case It's not wrong for use as an RGB working space. A zero black point for an RGB working space is exactly right. It allows the editing software to add RGB colors properly without first converting to XYZ. > - The media white of sRGB *profile* is D65 and should be D50 > - The chromatic adaptation tag is not there, so there is no way to > discount the illuminant. V2 color management manages to do chromatic adaptation using sRGB profiles with a D65 white point tag quite nicely. It was the V4 specs that messed that up. > - it is a V2 profile, not use perceptual reference medium gamut PRMG. > - etc. Is the PRMG of any use for an RGB working space profile? Certainly it has a use when outputting already edited images for printing/display (there is also the alternative of doing your own gamut mapping). But for actual image editing? > > ICC has addressed some of those issues: > http://color.org/srgbprofiles.xalter > > But still, gamut mapping may be improved and I would like to see a > floating point elements version, for example. Your wonderful unbounded profile conversions already work with RGB working space "simplified display" matrix profiles, and at least GIMP and darktable are already taking advantage of the possibilities. > Adobe RGB profile is almost same. I'm not sure what the context is. > > >> Do phrases like "legacy matrix-shaper profiles" and "only work with that >> very old profiles" - mean that using standard RGB working spaces (and >> linear gamma versions of same) is outdated or substandard for RGB >> digital painting or image editing, such as one might do using Krita, >> GIMP, or PhotoShop? > > sRGB space is an spec: IEC 61966-2-1, This is not outdated. > > The profile "sRGB Color Space Profile.icm" is a very old profile with > many problems. This is outdated. > > Linear versions are just different color spaces. The spec does not talk > about "linear sRGB" there is no such thing. This is something that uses > D65 as white point, gamma 1 and Rec709 as primaries, but this is not sRGB. I don't think anyone confuses "linear gamma sRGB" with the actual sRGB color space specs. At least I hope not. It gets confusing trying to qualify every word in every sentence to avoid confusing people. In common parlance, people do talk about "linear gamma sRGB profile" and hopefully everyone understands that what is meant is an ICC profile made with the D50-adapted sRGB primaries and the linear gamma TRC, and similarly with linear gamma ProPhotoRGB and linear gamma Rec.2020 and so on. I'm not sure where room for confusion might be. Is there a preferred phrase? "The profile you get when you make a profile that has the sRGB primaries and the linear gamma TRC" seems a bit cumbersome, and "linear gamma sRGB" seems pretty self-explanatory. > >> is there a way to retrieve these parameters from the profile and then >> use them to generate a point curve approximation that can be plotted? > > > lcms has those functions to do that: > > cmsGetToneCurveEstimatedTableEntries(const cmsToneCurve* t); > cmsGetToneCurveEstimatedTable(const cmsToneCurve* t); Thanks! Best, Elle ------------------------------------------------------------------------------ _______________________________________________ Lcms-user mailing list Lcms-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lcms-user
