Video series

[shijan]

It is also interesting that if look inside ICC profile info, in most cases you may see Perceptual tag. But in reality as we can see Relative Colorimetric transformation used.

[shijan]

If you make or download a large-gamut printer profile, then convert an image from a small gamut such as sRGB, the difference between RC and perceptual is clearly visible, and BPC works with both as well. My understanding is that the rendering intents are included in the profile itself, and it is printer profiles that have them both. Working space profiles, and I believe even display profiles, are evidently only able to use RC. So you just have to know that when you convert an image between working spaces.

Yes, in case of Printer ICC profile i can see difference in image by naked eye when i change Perceptual/Relative in Convert ICC profile settings window.
But same as in earlier example - it only shrink colors from larger color space to smaller. But do not expand colors from smaller color space to larger. Not sure if it should or should not work like this. I always thought that Perceptual should attempt to fill all possible color space (expand or "upscale" colors from smaller space to larger)

[russellcottrell]

RC
source too big: gamut overflow gets compressed and the rest is unchanged; relationships may change
source too small: nothing changes

Perceptual
source too big: entire gamut gets compressed; relationships do not change
source too small: entire gamut gets stretched; relationships do not change

[shijan]

Perceptual
source too small: entire gamut gets stretched; relationships do not change

Have no idea why it don't gets stretched in my example. Maybe some global limitation in color management? Or bug?
I also experiment with Perceptual/Relative setting combinations in both source and destination ICC profiles and also tested Apple CMS and Little CMS. Same result in all combinations - small color space remains unchanged and not stretched to larger size.

For compare in DaVinci Resolve same color chart normally fully transforms from smaller to larger color space, as you may see on scopes visualization:

[shijan]

So i checked some other apps (darktable (LCMS), Iridient Developer (LCMS), Photoshop (Adobe CMS), 3D LUT Maker Perceptual LUT (ArgyllCMS)). They all work more-less the same in Perceptual mode and don't stretch sRGB colors to larger color space.
DaVinci Resolve have few different color management modes and one of them "Simple" also don't stretch sRGB colors to larger space.

[russellcottrell]

I am getting similar results using ArgyllCMS to view the gamuts of Grainger rainbows created in sRGB and ProPhoto and converted to a printer profile with both perceptual and RC. The ProPhoto image gamuts both shrink down to the printer profile gamut as expected (can’t tell how they were shrunk, by local or overall compression). But the perceptual gamuts are identical when I would expect them to be different. The images themselves do not look very different.

Grainger rainbow on sRGB (identical)

GraingerOnsRGB.jpg

sRGB on the printer profile

sRGBOnPrinter.jpg

Grainger rainbow converted to printer profile, RC

RCOnPrinter.jpg

Grainger rainbow converted to printer profile, perceptual

perceptualOnPrinter.jpg

[Martin Huber]

I am a bit late and my knowledge of the internals of color management is limited.

I am working on a video(s) about color management. Neither rendering intent nor blackpoint compensation has any effect when converting between working spaces (ProPhoto to sRGB, for example). There is little written about this but my understanding is, that other program uses relative colorimetric for all such conversions regardless of what is selected. In PL there is no apparent change when converting from a small space (sRGB) to a large one (ProPhoto), so my guess is that it uses RC as well. Is this correct?

ICC profiles define how a color is converted to and from the so-called profile connection space (PCS). The PCS can be XYZ or Lab. So in theory, each color conversion converts the color from the source color space to the PCS and from there to the destination color space

Display profiles just define curves for the single color channels.
Other profiles (CMYK) use LUTs. There may be LUTs for the various rendering intents and LUTs for converting to the PCS and from the PCS, so one of those profiles may have up to 6 LUTs. This is the reason why some profiles can not be a source or a destination color space (the corresponding LUTs are missing) or give wrong results in some situations (the corresponding LUTs are wrong).

If two display profiles are involved, the rendering intent usually has no effect because with display profiles the conversion is unambiguous.

Detailed information can be found at https://color.org.
The "ICC White Papers" section might be most interesting. There is also a document about black point compensation:
https://color.org/WP40-Black_Point_Comp ... -07-27.pdf

Martin

[russellcottrell]

Working space profiles also just have RGB curves. But to convert between different-sized profiles, like ProPhoto and sRGB, there has to be some sort of gamut mapping, and it does not appear that rendering intent has any effect, in any application. Here is a line from
https://www.adobe.com/digitalimag/pdfs/ ... roprep.pdf
“The rendering intent is the method by which the colors in the image will be compressed or clipped when separated to CMYK.” Taken literally, it means that rendering intent does not apply to working space transformations, only between RGB and CMYK. And the only examples anyone ever gives for rendering intent are to convert to or from a CMYK or printer profile. So, do we know the rendering intent used to convert between working space profiles? Is it alwys RC? (I realize this takes place in Little CMS or the operating system CMS, not PhotoLine per se.)

Even the documentation for jpgicc at
https://manpages.ubuntu.com/manpages/im ... icc.1.html ,
part of Little CMS, gives six conversion examples, but no working space to working space.

The printer profile I used is GPGFS13_CANipf6350_PSGP280n.gam, downloaded from Ilford. It has all six LUT’s. What surprised me above is that the sRGB image gamut looks like it was merely compressed, not expanded anywhere. Presumably the compression math was different for the different rendering intents, but you can’t see it just by looking at the gamut.

p.s. I have an X-Rite system, and it can make display profiles with either curves or LUT’s.

[russellcottrell]

I think I realize the answer, don’t know why it took so long. Martin, if you replace “display profile” with “working space profile” in your post, it all makes sense. Colors are converted from one working space to another by a simple matrix transformation to and from XYZ, so one whole gamut is just mapped to the other 1:1. That’s why if you take the sRGB Grainger rainbow and convert it to ProPhoto, the gamut expands to fill ProPhoto exactly. What was new to me is that rendering intents evidently do not expand anything, just compress, in two different ways for perceptual and RC.

[bruce1951]

I'm commenting here way above my pay rate!! But it may help to have a look at QImage. They seem to have the colour questions sorted with their software. Maybe worth a question on their forum. https://ddisoftware.com/tech/qimage-ultimate/

[russellcottrell]

Well, I found this:
https://ddisoftware.com/tech/articles/j ... g-intents/
The part about a “smart” CMM was interesting. I always assumed that there was some sort of mind in there that mapped the gamut of the image to the destination profile; but it is an all-or-none affair, evidently built into the LUT’s.

This is where a gamut warning function would come in handy. GIMP has it, fyi. It is also available in Little CMS.

[Martin Huber]

Working space profiles also just have RGB curves.

I don't know the term "working space profile".

There are several types of ICC profiles. The most prominent ones are (these are the official names):
- display profiles
- output profiles
- colorspace conversion profiles

sRGB, Adobe RGB and ProPhoto are display profiles. The usual CMYK profiles are output profiles.

Each of these profile types has certain requirements. For example, display profiles use single channel curves and output profiles use multidimensional lookup tables.

The printer profile I used is GPGFS13_CANipf6350_PSGP280n.gam, downloaded from Ilford. It has all six LUT’s. What surprised me above is that the sRGB image gamut looks like it was merely compressed, not expanded anywhere.

I would expect that. Rendering intents are there to handle colors out of gamut. If there are no colors out of gamut, there is no need to interfere.

Martin

[Hoogo]

I was using the name "Display profile" for profiles for monitors and screens, and they have somehting like 'LG, Dell, Acer..." in their name.

And for me "Working space profiles" are sRGB, AdobeRGB... and other device independent profiles that tell how to interpret some RGB values. They have the useful feature that same RGB values make grey. Even though monitors often have sRGB, these are not "Display profiles" (how I used to use the word "Display").

[gennadiy]

And for me "Working space profiles" are sRGB, AdobeRGB... and other device independent profiles that tell how to interpret some RGB values.

In my opinion, "Working space profile" this is just a tag that is written in the image file.

[betotex]

This is great! Very welldone videos. Learned a lot! Maybe some Liquify videos? I´m just crazy about this tool on any software. Keep on!