Visualising Perceptual and Relative Rendering Intents

CHROMiX ColorThink
1

Hi all,

I’m a newbie to ColorThink Pro and am excited by its potential in helping me as an educator in the classroom to better explain Colour Management theory - a picture is worth a thousand words, right?!

I’ve been having the below discussion with Pat from Tech Support and he suggested I take it to the forum as it may be helpful to others so I’ll try to summarise what I’m trying to achieve below and then leave it open for others to chime in if they choose:

#############################################

Basically, I have an image I want to use for a class exercise which I think demonstrates the power of rendering intents within colour management quite nicely. I’d love to show the image here in this post (for clarity) but can’t see how to do it apart from a link to it from somewhere else, sorry. The image is basically a stylised picture of Lana Del Rey’s face (all solid colours) with a floral pattern running through the background and the skin areas. It was created in Photoshop as RGB. There’s essentially only 5 colours used (though the anti-aliased edges produce a lot more colours in the total image):

  1. 27.192.38 (dark green background)
  2. 15.222.30 (bright green pattern in background)
  3. 213.255.69 (olive skin)
  4. 160.217.1 (darker skin pattern)
  5. 0.0.0 (Black face contour and hair)

When the image is viewed in RGB, the colours are clearly visually different from each other, but when you print them to a CMYK device, using Relative rendering intent, the two greens almost sit on top of each other at the edge of the CMYK gamut and the results are that the floral pattern in the background almost completely disappears - a great teaching moment then arises! ;-D

I then want to use ColorThink Pro to import the image and compare it against the AdobeRGB source profile and the desitination Xerox Color 1000 Press profile. I then want to use the vectors function to show how the colours will move when the various rendering intents are set.

However, I’m not quite seeing what I’m expecting. According to my understanding of how the Relative and Perceptual intents work:

With Relative RI, out of gamut colours (eg. RGB) are supposed to stop (clipped) at the outer edge of the smaller gamut (eg. CMYK) and inner gamut colours move according to the white point. I’m seeing this nicely in the 3D grapher.

Regarding Perceptual, this intent maps the whole source gamut to fit the destination gamut. This means that even in-gamut colours can be moved closer to the neutral point and become less saturated as the outer edges of the source gamut fits to the destination gamut edge (I kinda think of it as a balloon being shrunk down to the smaller size).

So, considering this with regards to my training image: The #2 RGB green colour (15.222.30) is not fully saturated so it’s a fair distance in from the edge of the AdobeRGB gamut. When using Perceptual RI, this colour should then be mapped INSIDE the destination CMYK gamut, not to the edge of the CMYK gamut, right?

The CTPro 3D grapher is showing the colours stopping at the edge of the CMYK gamut, which doesn’t seem right.

So either my RI theory is wrong or the CTPro 3D graph is wrong… Maybe Im looking at it too much in terms of the shapes of the gamuts, instead of actual Lab values when colours convert from one gamut to another…

Either way, I’d love to know what’s going on so I can use this image in my teaching exercise and properly explain how it’s all working to my students.

Your help is MOST welcomed and appreciated! :smiley:

2

Hi Brett,

Welcome to the forums, and thanks for posting this. This is a great place for questions.

I’ve gone ahead and brought the image you sent me into the Worksheet, so we could apply the sRGB profile to it, click on each of the 5 colors with the eye dropper, and find the Lab values for these colors. (Note that the RGB numbers you supplied here don’t actually relate to colors until we apply a profile to them.) I can also save this out as a color list file, which sometimes makes it easer to handle than an image. In your case, this 5-color colorlist will work identically to your 5-color image.

When I bring these 5 “Lana” colors into the Grapher, I’ll choose Plot as Vectors, and make your Fiery 1000 the destination profile. I’ll also bring in your profile into the Grapher and switch the surface to “smooth” & turn the opacity down so that it’s easy to see through the gamut at anything the vectors are doing.

I can look at this setup from a few different angles and I confirm what you are seeing. The perceptual rendering intent does not show the colors being mapped to the inside of the profile gamut. The destination points are desaturated slightly because the location of the points is slightly closer to the neutral axis of the profile.

So, what does this tell us? You have a profile that moves these particular out of gamut colors into gamut in this particular area when using the perceptual RI. ColorThink is showing you what’s going on.

Let’s try another tactic. Choose the plus button in the Grapher to search for the Stunt files folder inside “Favorites.” Bring in the “Lab cloud” file, change it to vectors and choose the Fiery 1000 profile again as the destination. This will show you what an even sampling of all the colors in Lab will do when you run them through your profile. It’s a pretty confusing mixture of vector lines, but as you zoom in you can still see that many of the vectors end at the boundary of the profile gamut. To make it clearer, I’ll turn up the opacity slider and zoom in to take a look “inside the skin” of the gamut. Most of the vectors end at the gamut boundary and there is very little going on inside.

This may not be what we expect, but this is why you have ColorThink. You get to find out exactly what a profile is doing to your colors, rather than just guess. Does anybody know if EFI has some kind of secret sauce to their profile building whereby they can make a perceptual rendering intent that is smart about the gamut boundary and avoids desaturating in-gamut colors? I’d say we just stumbled upon evidence of a pretty cool feature of EFI profile building. Other profiles will certainly give you the desaturation that you’d expect from the perceptual rendering intent.