Pages: [1]   Go Down

Author Topic: Profile quality check - why deviation on primary colors?  (Read 2585 times)

salzrat

  • Newbie
  • *
  • Offline Offline
  • Posts: 5
Profile quality check - why deviation on primary colors?
« on: March 01, 2011, 11:11:27 am »

I'm trying to get to grips with color calibration and profiling. As a first exercise, I'm trying to calibrate/profile my crappy 15'' laptop display using an i1pro (Rev B).

Now something puzzles me: When I do a calibration to D65, sRGB-Gamma, and create a profile for that, obviously only the grey ramp can be calibrated using the graphics card LUT.

However, to my understanding, when evaluating the profile quality using profiling software (basiccolor or icolor), for a matrix profile, shouldn't at least the primary colors have a very low deltaE? What I'm seeing is very high deltaE (5-6). White (255 255 255) has deltaE =0, and the whole grey axis looks more or less ok. So why are the primaries so far off?

As far as I understand, a matrix profile stores the primary colors (R,G,B) in XYZ-coords, and a correction table for different luminances for each primary. But the color (255,0,0) should map exactly to the color of the first column of the matrix, since that's just the XYZ of the primary color. So why is there any deviation at all?

Further, is there any software that displays the profile along with the values measured for the profile validation (best in a 2D view), so I could understand where the deviation would come from? A plot of the gamut + the 3 primaries + the measurements for the three primaries would be great...

Thanks!
Logged

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: Profile quality check - why deviation on primary colors?
« Reply #1 on: March 01, 2011, 12:00:11 pm »

Quote
I'm trying to calibrate/profile my crappy 15'' laptop display

That should give you a clue.

Have you actually visually compared the appearance of laptop RGB primaries to the primaries on a dedicated high quality wide gamut display? It can vary greatly.

Besides the electronics and quality of laptop panels are designed to preserve battery life. Just one slight voltage inflection can throw off a colorimeter which are sensitive enough to see into blacks.

IPS panels on dedicated displays draw a lot more juice to produce their color fidelity so they aren't built into laptops, so you're left with a TN panel of questionable build and quality control standards.

The MacBook Pro's reviewed here seem to indicate there are exceptions...

http://www.robgalbraith.com/bins/multi_page.asp?cid=7-10041-10146
Logged

salzrat

  • Newbie
  • *
  • Offline Offline
  • Posts: 5
Re: Profile quality check - why deviation on primary colors?
« Reply #2 on: March 01, 2011, 03:30:13 pm »

Yes, I know that the screen is crappy, and I can visualize the gamut and it is really quite small (66% sRGB). Still the primaries are consistent in their color. So even if they are not even close to sRGB primaries, a profile should still be able to characterize the screen well, at least the primaries themselves. Note that to get the sRGB white, the graphics card LUT has to be adjusted quite strongly (the native white is around 5000), but still, with the R,G,B primaries the profile should be able to get a good match, or am I missing something (what?)?
Logged

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: Profile quality check - why deviation on primary colors?
« Reply #3 on: March 01, 2011, 04:16:02 pm »

Can you describe what you mean by the look/consistency of the "primaries"?

66% of sRGB?! 5000K native WP probably insures that. Is the native WP neutral looking as is or does it look overly warm? If it looks fine, calibrate/profile using native WP. There's only so much a LUT can correct for without both adjusting for neutrality AND low gamut primaries within the matrix part of the profile. The less curves applied to the video card the better off you'll be.

What are you wanting the colormeter to do? It's not a profile's function to give pleasing colors. It's suppose to write an accurate characterization of your laptop screen for CM apps to adjust their previews by. If the math written in the profile says how REALLY crappy your laptop is then you have no choice but to go by what the math says. It may not be to your liking but that's the price you pay in a color managed workflow.

I hope you're not planning on editing images on that laptop. I'ld strongly advise against it.

Something else just came to mind. Make sure you don't select v4 profile generation in your calibration software, stick with v2.

 
« Last Edit: March 01, 2011, 04:22:13 pm by tlooknbill »
Logged

salzrat

  • Newbie
  • *
  • Offline Offline
  • Posts: 5
Re: Profile quality check - why deviation on primary colors?
« Reply #4 on: March 01, 2011, 10:15:24 pm »

Thank you for indulging the discussion!

My primary goal is still to understand the functioning of the ICC profiles.

I know that I am better of calibrating to native WP, and that looks much better than sRGB. However, I still want to understand the high deviations that I get when validating the sRGB-calibration+profile, just as a matter of principle.

Again:
- I calibrate the screen (via video LUT) to sRGB WP+Gamma.
- I create a matrix/trc-profile
- I validate that profile using the same software that created the profile (without even changing the position of the sensor), and get >5 deltaE for the primaries.

This I simply don't get.

Let's say my calibration resulted in an R-value of XYZ_1. Then the profiling process should measure a patch with (255,0,0) and get XYZ_1 as a result, and thus use this as the first column of the ICC-matrix. On validation, I would expect the software to display (255,0,0), measure again XYZ_1 (or something quite similar, depending on the repeatability of the i1pro). On the other path, the software multiplies (255,0,0) with the matrix in the ICC profile, which should again give XYZ_1, so there should not be any (or only a minimal) deviation. Even if XYZ_1 has nothing at all to do with an sRGB red! See what I mean?

Logged

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: Profile quality check - why deviation on primary colors?
« Reply #5 on: March 02, 2011, 09:34:36 pm »

I see what you're saying, but I can only guess at the cause.

I've never calibrated any display that had a gamut size 66% of sRGB. Since most colorimeters in general have RGB filters tuned to expect sRGB, I'm guessing it's seeing too many color holes to fill and/or patch up relying on LUT curves and matrices. There's only so much a math and inexpensive hardware can do to compensate the electronics of all displays.

It also could be voltage inflection in such a low powered device as your laptop. I'm assuming you're not calibrating the laptop relying only on its battery power.

Back in the day calibrating cheap CRT's I noticed color patches during the profiling process actually fluctuate with a quivery kind of behavior slightly changing in intensity and hue. I thought to myself that can't be good.

I really never did any validation measuring of any of my displays. If a standard color target viewed in Photoshop looked as before after calibration I went by that. I'm not really concerned with Delta E numbers.

Sorry I can't help you with this. Maybe someone more knowledgeable about validation software can explain better and come up with a more viable answer for you.
Logged

salzrat

  • Newbie
  • *
  • Offline Offline
  • Posts: 5
Re: Profile quality check - why deviation on primary colors?
« Reply #6 on: March 03, 2011, 03:54:49 am »

Ok, thanks.

RGB filters can't have anything to do with this, the i1pro is a Spectrophotometer :)

And the colors on the display look very stable, I don't believe in a flickering effect.

Must be something in the ICC calculation which I don't get...
Logged

Czornyj

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1948
    • zarzadzaniebarwa.pl
Re: Profile quality check - why deviation on primary colors?
« Reply #7 on: March 03, 2011, 04:06:21 am »

Maybe you just didn't assign and activate the display profile? Close basICColor display, check if the profile is properly assigned in Color Management settings and the correction is loaded into the graphics card LUT, then start basICColor again and perform the validation once again.
Logged
Marcin Kałuża | [URL=http://zarzadzaniebarwa

howardm

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1984
Re: Profile quality check - why deviation on primary colors?
« Reply #8 on: March 03, 2011, 06:45:51 am »

It's been my experience that BasICColor isn't kidding when they tell you to quit/restart the application in order to do a validation of a newly created profile.  I also saw high dE's

salzrat

  • Newbie
  • *
  • Offline Offline
  • Posts: 5
Re: Profile quality check - why deviation on primary colors?
« Reply #9 on: March 03, 2011, 04:29:10 pm »

I'm positive that the LUTs were really changed (I checked with x-rites calibration tester, a very handy tool)...
Logged
Pages: [1]   Go Up