Pages: [1]   Go Down

Author Topic: Delta-E, when is it too high?  (Read 16577 times)

Hendrik

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 93
    • http://www.hendrik-fotografie.nl
Delta-E, when is it too high?
« on: July 08, 2006, 03:26:22 pm »

When are the delta-E values too high? The last couple of times I seem to get worsening of the blues with now two values above 3 (3,16 as the highest). The mean delta-E is 1,57

I remember something that a delta-E above 3 is practically relevant while doing your editing. So were do you draw the line and how can I optimize my display a little longer?

(The strange part is that the greens are the best, while I reduced its amount by 50% (red 100%, blue 80%).

btw, I use Monaco Optic XR Pro
Logged

Jack Flesher

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2592
    • www.getdpi.com
Delta-E, when is it too high?
« Reply #1 on: July 08, 2006, 04:15:36 pm »

Quote
When are the delta-E values too high? The last couple of times I seem to get worsening of the blues with now two values above 3 (3,16 as the highest). The mean delta-E is 1,57

I remember something that a delta-E above 3 is practically relevant while doing your editing. So were do you draw the line and how can I optimize my display a little longer?

(The strange part is that the greens are the best, while I reduced its amount by 50% (red 100%, blue 80%).

btw, I use Monaco Optic XR Pro
[a href=\"index.php?act=findpost&pid=70088\"][{POST_SNAPBACK}][/a]

I think anything under 2 is excellent, over 3.3 and you are in not-good territory.  3.16 is marginal...
« Last Edit: July 08, 2006, 04:16:40 pm by Jack Flesher »
Logged
Jack
[url=http://forum.getdpi.com/forum/

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Delta-E, when is it too high?
« Reply #2 on: July 08, 2006, 05:58:17 pm »

Too high based on what? If you're using the same instrument to measure the device and produce the delta, it's kind of a figure there just for entertainment purposes. IOW, if you have a reference grade instrument with a known accuracy and use that to compare the data from whatever other device you used to calibrate the display, you can gain useful delta's. Otherwise, you're measuring everything with the same yardstick which isn't very useful.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Serge Cashman

  • Full Member
  • ***
  • Offline Offline
  • Posts: 200
Delta-E, when is it too high?
« Reply #3 on: July 08, 2006, 08:04:38 pm »

It does mean that the result is different from what was intended by the calibration software though. It indicates that it was not able to accurately reach the calibration targets. Which is probably not a big deal from a user point of view, but still.
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Delta-E, when is it too high?
« Reply #4 on: July 08, 2006, 09:07:13 pm »

Quote
It does mean that the result is different from what was intended by the calibration software though. It indicates that it was not able to accurately reach the calibration targets. Which is probably not a big deal from a user point of view, but still.
[a href=\"index.php?act=findpost&pid=70111\"][{POST_SNAPBACK}][/a]

It simply says that the measured data and reference data are not the same based on this instrument. But that doesn't tell you that the resulting data is correct or not nor does it tell you what was measured was accurate or not. It's just different but not necessary wrong.

If the instrument is off a deltaE of 2 due to its inability to measure dark tones on the display, the resulting deltaE the software provides can't know this (because the instrument used to provide the deltaE can't "see" or measure these dark tones.

If I have a 12 inch measuring stick that is inaccurate by half an inch, using it to re-measure something doesn't tell me very much. If I use a reference grade tape measure that is accurate to 1/1000th of an inch and it tells me I'm off a certain amount, that's useful info.

And what's the accuracy of multiple measurements of a device that's off half an inch? IOW, if one time it tells me I measured 12 inches but the next time it said the same unit measured was 12.3 inches, what am I to make of this delta?
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Hendrik

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 93
    • http://www.hendrik-fotografie.nl
Delta-E, when is it too high?
« Reply #5 on: July 09, 2006, 05:32:13 am »

…but what’s then the purpose of evaluation tests and graphing the trend?

There must be something useful in it besides entertainment, or isn’t it?
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Delta-E, when is it too high?
« Reply #6 on: July 09, 2006, 10:04:45 am »

Quote
…but what’s then the purpose of evaluation tests and graphing the trend?

There must be something useful in it besides entertainment, or isn’t it?
[a href=\"index.php?act=findpost&pid=70135\"][{POST_SNAPBACK}][/a]

If the instrument is consistent (and modern Colorimeters and Spectrophotometer's are) then trending is a more useful report which in the end just tells you how important it is to calibrate your display (and to some degree how often). But it doesn't tell you how accurate the process is from a standard baseline.

Again, what is the deltaE based on? Do you have a known reference baseline to use?
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Serge Cashman

  • Full Member
  • ***
  • Offline Offline
  • Posts: 200
Delta-E, when is it too high?
« Reply #7 on: July 09, 2006, 12:30:00 pm »

Since it's a Monaco XR one has to assume it's as accurate as consumer-grade equipement gets.

The Delta E that Monaco software measures from what I can tell is the difference between patches of color displayed by the monitor and the colors that those patches would display if the monitor exactly matched the calibration targets.

One issue with calibration accuracy is how it is done. If you do RGB adjustments manually (which is usually recommended for CRTs but not recommended for 8 bit LCDs) then it may be a slight user error.  To compare manual results with LUTs-based results set your monitor to default (6500K or Normal or whatever is in the middle) and indicate to software you only have RGB presets. See if LUTs-based calibration will be more accurate.


Again - I don't think it's such a big deal because from what I understand what gets described in a profile are the actual measurements and not the target values, so information used by colormanaged applications should still be correct.
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Delta-E, when is it too high?
« Reply #8 on: July 09, 2006, 12:53:22 pm »

Quote
Since it's a Monaco XR one has to assume it's as accurate as consumer-grade equipement gets.

The Delta E that Monaco software measures from what I can tell is the difference between patches of color displayed by the monitor and the colors that those patches would display if the monitor exactly matched the calibration targets.

[a href=\"index.php?act=findpost&pid=70174\"][{POST_SNAPBACK}][/a]

That's an excellent device, yes.

But you're still missing the point I think. OK, so lets say the software has a reference LAB value it expects. Since the same device is used to measure it AND we don't know its accuracy, the deltaE between the reference and measured data is still not clear. If a $300 device says the deltaE between reference and measured is 3 but you pop a $20K spectroradiometer and measure the two and you're told the difference is a deltaE of 4.2, NOW you have an idea of the true differences between reference and measured data plus the deltaE of that $300 device from a reference grade instrument. Using the same device to measure the reference and produced color may tell you there's a difference but by how much and is that value accurate?


Quote
One issue with calibration accuracy is how it is done. If you do RGB adjustments manually (which is usually recommended for CRTs but not recommended for 8 bit LCDs) then it may be a slight user error.  To compare manual results with LUTs-based results set your monitor to default (6500K or Normal or whatever is in the middle) and indicate to software you only have RGB presets. See if LUTs-based calibration will be more accurate.

I'd prefer to look at the curves and see how linear they are do get some idea how much adjustment is happening at the graphic card level.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Serge Cashman

  • Full Member
  • ***
  • Offline Offline
  • Posts: 200
Delta-E, when is it too high?
« Reply #9 on: July 10, 2006, 02:59:22 pm »

I understand your point of not being able to judge the accuracy of a device by measuring the results with the same device.

I just assume the device is consistent enough. It shouldn't  vary by 3 delta E from it's own measurment, especially if that can be observed  consistently. It's measurement may well be off compared to a more accurate device but like you said there's no way to tell without a reference instrument. But according to the available device measurements the calibration targets are not accurately achieved.

So in my opinion the initial question is a valid one - why the results are not as close to the target as they used to be? I think the reason might be the manual RGB adjustments.
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Delta-E, when is it too high?
« Reply #10 on: July 10, 2006, 05:03:36 pm »

Quote
It shouldn't  vary by 3 delta E from it's own measurment, especially if that can be observed  consistently.[a href=\"index.php?act=findpost&pid=70272\"][{POST_SNAPBACK}][/a]

In very dark tones (very though to measure) I don't know that's not impossible. One of the reasons the Sony Artisan was so good was the mating of the device to the actual display and the cost to build a special colorimeter that had very, very good dark performance.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

Hendrik

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 93
    • http://www.hendrik-fotografie.nl
Delta-E, when is it too high?
« Reply #11 on: July 11, 2006, 05:32:40 pm »

I didn’t saw the large delta-E values in the very dark tones, but in the bright blue tones. With calibration, I tried to be close to the desired 6500K, but didn’t want to lower the green channel lower then 50% (see topic-start). I did get a very good match (approximately 60k difference).

Maybe its user error, but I fail to see how I can get a better results, …if it’s needed of course.
Logged

Serge Cashman

  • Full Member
  • ***
  • Offline Offline
  • Posts: 200
Delta-E, when is it too high?
« Reply #12 on: July 12, 2006, 06:38:47 pm »

Quote
Maybe its user error, but I fail to see how I can get a better results, …if it’s needed of course.
[a href=\"index.php?act=findpost&pid=70387\"][{POST_SNAPBACK}][/a]

Try the LUTs-based calibration instead (by indicating to the software you don't have RGB buttons, only presets).  Your delta Es may be either lower or higher, but at least you'll get something to compare your results to. If you use an 8-bit LCD and get lower Delta Es you might as well keep calibrating that way.

I don't think it's crucial for any practical purposes (if you only use a single monitor) but it's still good to know why the error occurs.
« Last Edit: July 12, 2006, 06:42:05 pm by Serge Cashman »
Logged

ato

  • Newbie
  • *
  • Offline Offline
  • Posts: 29
Delta-E, when is it too high?
« Reply #13 on: July 13, 2006, 12:23:37 am »

buy a good CRT can fix your problem

this is my CRT result,anyone better than me?
Logged

Serge Cashman

  • Full Member
  • ***
  • Offline Offline
  • Posts: 200
Delta-E, when is it too high?
« Reply #14 on: July 13, 2006, 01:35:16 am »

Oh, so now you are HAPPY with your results?  Cause your delta E is smaller than Hendrik's?

Delta E doesn't matter, you know...  
« Last Edit: July 13, 2006, 01:36:15 am by Serge Cashman »
Logged

ato

  • Newbie
  • *
  • Offline Offline
  • Posts: 29
Delta-E, when is it too high?
« Reply #15 on: July 13, 2006, 05:23:17 am »

Quote
Oh, so now you are HAPPY with your results?  Cause your delta E is smaller than Hendrik's?

Delta E doesn't matter, you know... 
[a href=\"index.php?act=findpost&pid=70538\"][{POST_SNAPBACK}][/a]


yew,i happy that i'm using a good monitor.

and you say delta E doesn't a matter?
but i want to ask.....why most professional display had mention how good their delta E value?
barco ref montior,ACD.....etc
um...may be 3dmark better for me to check my monitor
Logged

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Delta-E, when is it too high?
« Reply #16 on: July 13, 2006, 01:33:28 pm »

Being someone that doesn't even use Delta E numbers for any matching, I can take an noneducated guess based on shear observation why exact Delta E's don't matter.

Adaptation. If my eyes can notice this without any calibration intervention, it pretty much makes Delta E numbers seem like measuring a moving target.

Just my opinion.
Logged

Serge Cashman

  • Full Member
  • ***
  • Offline Offline
  • Posts: 200
Delta-E, when is it too high?
« Reply #17 on: July 13, 2006, 06:27:21 pm »

I was mostly kidding...

Well, Delta E is just a difference between two colors (calculated as distance between 2 points in 3D space). For it to make sense it has to be defined which colors you are comparing. In this case you are comparing colors achieved after calibration to the calibration targets. If they are slightly off it's not such a big deal since the real measurments are used for gamut conversions anyway.

What Delta Es are used in online display reviews is a complete mystery to me. They just measure a bunch of patches and compare them to some other bunch of patches without telling you what the hell they are comparing them to... Essentially they compare uncalibrated monitor output to... something. All reviews I read don't tell you what it is.
« Last Edit: July 13, 2006, 06:27:49 pm by Serge Cashman »
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Delta-E, when is it too high?
« Reply #18 on: July 13, 2006, 06:38:14 pm »

Where deltaE is useful would be the following: Measuring the SAME color with the same device over the surface of the display. Be nice to know for example that a gray over the entire display is very, very close from center to edges and so forth (in reality, that's usually not the case. The old Barco Reference V could do a 25 quadrant purity adjustment using it's colorimeter. Cool and one reason even in the old days, you'd pay $5000 for one and see a deltaE of 1 over the entire glass).

Its also useful for trending. So the same instrument measures the same patches today, in a week and then in a month. You see how stable the calibration is and how the display might be aging.

Most important, discussing deltaE (like Metamerism) at a party impresses some chicks <g>. Of course, you'd be hard pressed to want to go too much farther with such a lady considering the non existent custom RGB color spaces named after woman or the lack of female color geeks (this is one chauvinist business!).
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".
Pages: [1]   Go Up