Mapping Gap?

I think i1p’s mappings from AdobeRGB are not using as much of my printer’s gamut as pm5’s colorful mapping. In ColorThink, compare 3D graph of i1p and pm5 device gamuts for same paper/inking combo to confirm that they match almost perfectly. Start with Bruce Lindbloom’s image that has a pixel for each of the 16 million 8-bit rgb colors, in Photoshop assign it AdobeRGB then convert one copy to printer space using i1p mappings and another copy using pm5 mappings. Bring the converted files into ColorThink 3D graph and see where the pixels reside inside the printer space. While pm5’s colorful mapping nearly fills the printer’s space, i1p’s mappings leave a significant gap on the outer edge of yellow/orange.

Am I looking at this the right way? Any explanation?


  • Brad[/img]

Any chance of sharing the screen shot with all of us? Maybe profiles also?

Does the same thing happen when you do the entire process inside ColorThink, leaving PS out of the equation?

Here are links to screen shots from ColorThink 3D graphs:

pm5: … hotostream

i1p: … hotostream

I don’t yet know how to do the entire process inside ColorThink. The worksheet still intimidates me. I’ll work on it. Though I like doing the conversions in Photoshop to see the actual device RGB values that I’m sending to the printer.

No no, don’t worry about the worksheet if you’re not comfortable with it. I was referring to graphing the image and then clicking on ‘Plot As: Vectors’ and simply dragging the profile into the area below, where it says ‘Destination’. This way you can view either the amount the colours change in deltaE or you can view them as true colours.

I find this such an extraordinarily useful technique, I am constantly using it. It gives awesome detail, especially once you play around with the sizes of the points and their shape (I find sphere the best) and I’ve worked out a few other little ‘tricks’ that make it even better and even easier to ‘read’ what’s happening.

Give that a try and I’m sure you will have a very useful new tool to add to your arsenal. :smiley:

Aaron, Thanks for the suggestion on using vectors inside ColorThink. I get the same positional results as my Photoshop conversion method, but it’s faster, easier, and more informative to see the vectors.

I believe what I’m experiencing is that the yellow/orange outer edge of my 9900 gamut is actually beyond the AdobeRGB space.

The pm5 mapping is being more aggressive at keeping these colors near the outer edge of the AdobeRGB/device gamut intersection, whereas the i1p mapping is pulling them in a little farther. If I start with the image assigned Bruce Lindbloom’s BetaRGB, which is just slightly larger than AdobeRGB in all directions, then all the mappings perfectly fill the printer space.

Seems like I should start using BetaRGB (or ???) instead of AdobeRGB for my working space when color matching for art reproduction. Matching colors in the light yellow to light orange range has always been a challenge, which is why it freaked me out to see that specific part of my printer’s gamut going unused. Maybe using a slightly larger working space that fully encompasses the printer space will help make this matching easier.

I’ve thought about using ProPhotoRGB, but it seems too extreme given that the colors I’m dealing with are reflective light from oil, acrylic, and watercolor paintings. AdobeRGB seemed the right choice when I first started printing artwork on 7600.

Any other thoughts on working spaces? Suggestions welcomed.


  • Brad

What RI intent did you use for the conversions? That will also have a large effect on the gamut mapping. I’m guessing you might have used relative as you explained you’re doing art repro so likely perceptual is not what you’d want?

Regarding colour space, I really see no reason not to use ProPhoto these days, especially when you think about how things might be 5 years down the track and if you wanted to re-process these files because the technology is so much better, if you use a smallish space, you’ll be possiby severely limiting yourself in the future.

Also, there would definitely be many places where a 9900 could exceed the gamut of AdobeRGB, if printing on a non-matte paper. The only other colour space I’d really recommend would be those created by Joseph Holmes, with his DCam variations for digital photos and his other specialised colour (working) spaces. With those, I would definitely recommend to use a space only just larger than the image tones need because with his variations and the way he has created these profiles, the idea is to make as much use as possible of the profile gamut volume, unlike with ProPhoto, where there is a huge amount of wasted space which will never be able to be used.

If you’re not familiar with Joseph Holmes’ self-created profiles and profile variations, I very highly suggest you look into them as they could be just what the doctor ordered. Or what you’re looking for, whichever you prefer. :stuck_out_tongue:

I’ve always thought I SHOULD use Relative rendering intent, but every time I tried in the past (with Atkinson and various other profiles on 7600, then with ImagePrint on 7600 and 9800), I had problems with plugging shadows and subdued colors whenever I tried Relative on my test images, like the black point compensation was never quite right and the mappings weren’t even trying to use the full printer gamut, so I always gave up on Relative and focused my efforts on getting to know my perceptual mappings well and just working with them.

Also, the idea that Relative handles out of gamut colors by clipping them and clustering them on the edges of the printer’s color space always bothered me. Scaling them perceptually into the printer’s space, delineating as many unique colors as possible for smooth, pleasing gradients seems more important. Plus, the plugging shadows issue always dominated, making everything else moot.

Now, with 9900, my own spectro, ColorThink Pro, and both pm5 and i1p as mapping engines, I’m trying to delve deeper, and a renewed analysis of Relative versus Perceptual for my workflow is in order. But the very idea of clipping and clustering the out of gamut colors versus perceptually scaling them really bothers me. Am I looking at this correctly, or is there some concept that I’m missing?

The very easy to understand concept you’re missing, IMHO is that relative is most often used when the source gamut is either totally contained by the device gamut or just a little out of gamut for the destination device. Perceptual would mostly be used when, as you mention, there are numerous parts of the source gamut which exceed or greatly exceed the destination gamut.

In that way, if your destination fully encompasses your source, there’s no problem at all to use relative, as nothing would be clipped and also nothing would be changed around if it were already inside the destination gamut. Because even if the source fits completely inside the destination gamut, using perceptual will always, 100% of the time, end up with some colours/hues/tones/etc being moved around, as that is just how perceptual works.

One thing I completely forgot to ask previously is what app did you use to create the original printer profiling target? This is extremely important with i1p as it is (well) known that the i1p engine works much better with it’s own targets, created within i1p than it does with PMP5 targets.

So, I’m wondering if perhaps you used a target created in PMP5, therefore already crippling i1p’s mapping slightly, before you even start creating the profile or doing any conversions?

I have measurements for my different media using targets created by i1p AND targets created by pm5. Both sets have identical printer settings and slightly more than 2000 patches. So I can test i1p’s mapping when using its own targets versus when it has to use pm5 generated targets.

Thanks for your input on why to use relative versus why to use perceptual. I understand now and totally agree. After quite a bit of study and experimentation, I think relative colorimeteric is far superior for my purposes than perceptual. My printer spaces are close enough to my working spaces that the out of gamut color clipping issue I feared (versus perceptual scaling) is not really much of an issue. It’s extremely helpful to be able to see exactly what’s going on by using Colorthink. I’ve learned to use the worksheet.

I read Adobe’s paper on how their black point compensation works and that was helpful. Colorthink’s relative mappings don’t use black point compensation, so actually doing the conversions to printer space in Photoshop then viewing the mappings in Colorthink is useful. With bpc turned off, photoshop relative mappings are nearly identical to Colorthink’s mappings.

Almost all the original colors I’m capturing fit within AdobeRGB. I can test the questionable colors using my Eye One before I shoot, which is kind of fun, to see if they’re outside AdobeRGB. If so then I can use BetaRGB as my target working space rather than AdobeRGB, but this need is rare.

When I compare the media relative mappings of a 256 grayscale chart on five different papers in the worksheet, I see how the grayscale values get spread between the media black and media white points. When I look at those mappings, it makes me wish I had a “midpoint compensation” switch for the relative mappings so I could nail L50 to its absolute value and have the mappings adjust the tone curve above and below that as needed to accomplish an absolute midpoint.

I’ve tried to get a better understanding of absolute colorimetric mappings. I played with manipulating absolute mappings in printer space using photoshop–taking white to be paper white, creating my own black ramp, and seeing how the mapping of vibrant colors differed from a relative mapping both in photoshop and ColorThink. Not sure I really understand the nuances of absolute mappings yet. More study required.

I’m interested in gaining a better understanding of the media relative shifting of all colors when profile connection space white gets mapped to media white. I feel as if I’d like to have some control over the media relative color shift so I could keep more colors at their absolute mappings, yet still have a black compensation ramp, still have PCS white mapped to media white, have some volume of near neutrals get mapped to be neutral on the media, and have some volume of light tones, where media white has a distinct influence, get shifted relative to media white. Or maybe I’m totally crazy?

I haven’t tried anything other than the default “Profile White Point” setting in i1p yet. Seems interesting but not sure what it’s used for. Any experience with this setting?