AVS › AVS Forum › Display Devices › Display Calibration › Basic Guide to Color Calibration using a CMS (updated and enhanced)
New Posts  All Forums:Forum Nav:

Basic Guide to Color Calibration using a CMS (updated and enhanced) - Page 12

post #331 of 1810
10/17/07 at 5:18pm
Thread Starter 
I knew that this would provoke a spirited response. The dE difference varies according to the way you measure dE. Here are the averages:

dE(uv) 1976: 6.5
dE CIELAB 1976: 5.6
dE CIELAB 1994: 1.9

To put this in some perspective, I did a survey of several modern digital displays some time ago using dE CIELAB 1976 and the BEST performing display AFTER applying the included CMS revealed an average dE of 4.7. The worst was 32.5. The middling performers were in the mid-teens. Furthermore, a lot HD video is apparently mastered with monitors that use SD phosphors.

Given that reality, I guess all that I was saying that it is hard for me to get too worked up over the differences between SD and HD color spaces. I've looked at the numbers and I've seen nothing that would cause me to change my mind. When we get to a time when displays are a lot more color-accurate than they are now I'll certainly reevaluate.

On a side note, the one real profound difference between dE(uv) and dE CIELAB is how they treat red and blue (they seem to treat green similarly). The difference between:

0.630, 0.340, 0.212
and
0.640, 0.330, 0.213

is 5.8 in dE CIELAB, but 12.6 in dE(uv). For blue, the difference between

0.155, 0.070, 0.087
and
0.150, 0.060, 0.072

is 9.0 in dE CIELAB and 4.3 in dE(uv). It looks like dE(uv) perceptually weights differences in Lightness less than CIELAB.
Reply
Reply

AVS Top Picks

post #332 of 1810
10/17/07 at 7:14pm
I think if you can get the calibration down to the dE=5 or lower level in either system it's probably worth targeting the correct primaries because you don't want to add to that error. Even if you can only get dE's of 10 why add another 5-10 on top of it? I think that certainly holds true for unambiguous sources such as SD SMTPE-C. If you're uncertain about the source primaries then it's probably a wash. Also, better CMS options will eventually be available. A week ago I never thought it would be possible to get accurate primaries out of my panny display but now based on an HTPC solution in this thread I can watch DVD's in near-perfect color.

Before/After gamut's



dE(uv) RGBYCM=[4.7 5.5 4.1 4.2 10.3 7.5]
Reply
Reply
post #333 of 1810
10/17/07 at 8:07pm
Thread Starter 
Quote:
Originally Posted by alluringreality View Post

One thing I still don't really understand is the HD vs SD colorspace issue. When I measure DVE HD I get one set of Y values for color, but if I measure GetGray I get a second set of Y values for red, green, and blue.

I just measured using DVE HD (I finally found the color patterns. If anyone is interested they are in the 720p section under "Uniformity" of all places.), GetGray, and my Accupel signal generator. The Accupel and HD DVE were identical. GetGray was down a very small amount, about 4%. I attribute this to the fact that it was measuring a window instead of a full field. In other words, I was measuring absolute values, not color/white proportions, and there was only a 4% difference.
Reply
Reply
post #334 of 1810
10/17/07 at 8:16pm
Quote:
Originally Posted by zoyd View Post

dE(uv) from your spreadsheet between SMPTE-C and BT.709 RGBYCM is
12.6
6.7
4.3
0.9
5.7
9.8

Thanks zoyd - I have never actually sat down and worked it out - just lazy!

Red certainly is the difference I see that really annoys me when SD is stretched out to HD. Indeed I would have guessed the worst offenders around 10% perceptually off. Of course only a calibrator could get annoyed by that - when J6P has no clue greens are 60% off because the mfg thinks they should hire TedTurner as their Video Consultant instead of JKP....

Anyways those errors are right over my videophile edge of damn I need to get this calibrated as it is annoying vs. how much work is it to do that considering nobody pays me to calibrate MY TVs!. I was all set to follow that HTPC thread and give it a shot on colour correction - but my HTPC has died - spendy quiet powerful power supply took a hit.
I need to look at dH for Yellow though vs. my perception - I know I can see it as being off - but Yellows dL or dC I would find very hard. I don't mean SD or HD Yellows which are little different - I mean Yellow errors in general.

I think the perceptual differences in light sources vs. surfaces is why we need to use CIELUV for video. I fail to see how CIELAB tracks perception of light - it is specifically for the color of surfaces (the absorption of lite if you will) If you want to get into lighting the surfaces - that is a whole different section of color space science that nobody here has even talked about - but certainly would be applicable to PJ screen colourations (maybe - as I think the science applies more to ambient lighting more than strong focused direct illumination - though I have not studied this in depth)

Blue is very dark light and will bias your perception to minimize the errors - Yellow is very light and it takes greater swings in the lite before you see the errors - Red is more of a balance - just bright enough that we notice the smaller errors. Our eyes are logarithmic light sensors - but the light function is based on Y (not Yellow - the CIE Y) - which is dominated by greens with very little blue and some red. CIELAB is very different in that each XYZ is log scaled - not just the Y as in CIELUV (which in turn is used to scale the errors)

Just look at a correct SD vs. HD displayed gamut - and I think you would perceptually agree that CIELUV is correct making the red difference much worse than the blue difference, rather than CIELAB which has the blue difference as a bit worse than red.
Reply
Reply
post #335 of 1810
10/19/07 at 2:18pm
Thread Starter 
I have just substantially enhanced the original tutorial. It now includes instructions for setting Contrast, Brightness, and Sharpness.

I also included links to download a basic calibration DVD I put together that has all of the necessary patterns I talk about in the tutorial.
Reply
Reply
post #336 of 1810
10/19/07 at 2:54pm
tom:

when you say peek output do you mean the Y of 100% white?
Reply
Reply
post #337 of 1810
10/19/07 at 3:01pm
Many thanks Tom. An excellent tutorial now becomes complete.
I can't wait to try the techniques laid out on the tutorial on my new Sharp XV-Z21K

Regards,

kopa13
Reply
Reply
post #338 of 1810
10/20/07 at 7:54am
Hi Tom, thanks for new tips! I have a question to the contrast part.When i use HCFR , by peak output you mean the max. number i get, when i measure contrast with HCFR? Becose if i undersnd it right i would need to set contrast at maximum on on My TW1000 and still not get to 15 fL
Reply
Reply
post #339 of 1810
10/20/07 at 10:04am
Thread Starter 
Yes, peak output is Y from a 100% white input.

The SMPTE standard is 12fL. I specified 15 fL only because UHP lamps lose brightness fairly quickly. If you are getting less than 10 fL at peak output (remember, you need to take screen gain into account), then you need a new bulb.
Reply
Reply
post #340 of 1810
10/20/07 at 10:53am
I get 9 with contrast set to maximum, but with it i cant create a reasonable gamme curve. with gamma at 2.2 and contrast on 0 its around 6...i dont know if epson would accept, that they should give me new bulb, becose i have only 700 hrs on the lamp and they give 1500hrs or 3 year garanty on it. But i have severe shading on my pj and the right part of the picture is noticebly darker, so i wanted to send it to service center to be checked if they can do something with it enyway ...
Reply
Reply
post #341 of 1810
10/20/07 at 12:16pm
Thread Starter 
Quote:
Originally Posted by krasmuzik View Post

I think the perceptual differences in light sources vs. surfaces is why we need to use CIELUV for video. I fail to see how CIELAB tracks perception of light - it is specifically for the color of surfaces (the absorption of lite if you will) If you want to get into lighting the surfaces - that is a whole different section of color space science that nobody here has even talked about - but certainly would be applicable to PJ screen colourations (maybe - as I think the science applies more to ambient lighting more than strong focused direct illumination - though I have not studied this in depth)

Kras:

I done a little research on this and I can find no evidence that either color space is specially designed for light sources or surfaces (Indeed this source suggests that CIELUV was preferred in this study for determining the color of soil!).

Both color spaces were proposed in 1976 as an adaptation of the 1931 CIEXYZ color space in search of a specification that was more perceptually uniform. As the Wiki article on CIELUV puts it: "CIELUV and CIELAB were adopted simultaneously by the CIE when no clear consensus could be formed behind only one or the other of these two color spaces."

Both are intended to more accurately represent color as perceived by the human eye. There may have been historical trends wherein CIELUV has been used by the video industry and CIELAB has been used by the print industry, but I can find no reference that states that either standard is inherently unsuited for either. In fact, I've found several that state that they are equally suitable. For example, this source claims that they are only slightly different and which you use is merely a matter of "individual taste." This NIST source specifically recommends both for light sources due to their perceptual uniformity.

It seems to me that much more important than whether we use CIELAB or CIELUV is the question of the version of dE used and how to interpret it. The results for dE 1976 and dE 1994 (available only for CIELAB) are radically different, which at a minimum means that we require different standards for interpreting each. Is the minimal acceptable difference 5, 3, or 1 dE? It can't be the SAME standard for both, since the 1994 numbers seem much lower.

If you can point me to any reference that speaks to this, I'd really like to see it.
Reply
Reply
post #342 of 1810
10/20/07 at 12:50pm
My textbook references are listed in my spreadsheet (I also have other textbooks). I do not consider wikis or google to be references. The references I used will tell you that CIELAB dE intentionally made gross errors (of the type you get in video) less accurate in order to get less perceivable errors (but needed to ensure color surfaces are reproduced exactly) more accurate - so most certainly the different dE is a different tolerance curve - to the point CIELAB it is not even recommended for use over 5dE.

More importantly do the test yourself - does the various dE measures match what your own eyes are telling you if used as a percent? I predicted SD vs HD would be at worst ~10% based on my own eyes and only one dE standard agrees with that. Can you perceive the Red SD as being more off HD than the blue SD or vice versa (not subjective annoyance - just pure perceptual difference). I think even yourself said the CIELAB on the JVC RS1 Green seemed out of wack with perception.

After psychovisually testing CIELUV with customers (at shootouts and calibrations - specifically L*u*v* and LCH - not L*u'v') for years now to see if the charts agree with what they are seeing (vs. CIE xy charts which they never could see how it tracks - and neither could I which is why I switched) - I am convinced that CIELUV having dE be a stronger function of Y (since we perceive the gamma curved L*) is the reason it works so well to model light color perception severly impacted by luminance - very important in video when RGB is intentionally unbalanced in brightness.

Your NIST source simply said CIE1976 standardized on both as more perceptually more uniform than 1964 or 1931 versions. It did not say which of CIE1976 was preferred over the other for which application.
Reply
Reply
post #343 of 1810
10/20/07 at 2:13pm
Thread Starter 
Kras:

Other than the Wiki pointer, these were references to either official documents or academic papers. The mere fact that they are published on the Web doesn't make them less authoritative, so I don't really understand your comment about Google. Your spreadsheet references Bruce Lindbloom's web site, so I assume you treat his work as authoritative, despite the fact that it can be easily found by a Google search and despite the fact that he treats dE exclusively a function of CIELAB. It's Lindbloom's site that got me interested in this in the first place.

The NIST document in particular is specifically about light sources and it quotes the CIE recommendation of both standards without further comment. I would have thought that if CIELAB were completely unsuitable for light sources, they would not have done so.

There is no question that CIELUV is more perceptually uniform than what we get from standard xy charts. That's what the 1976 recommendations were all about. But that does not speak to the issue I raised.

I'll take a look at the sources you cite and report back. I don't want to continue this discussion any further until I read what you've read and access it along with other relevant material. I can't comment intelligently on what I haven't seen.

Quote:
Originally Posted by krasmuzik View Post

My textbook references are listed in my spreadsheet (I also have other textbooks). I do not consider wikis or google to be references. The references I used will tell you that CIELAB dE intentionally made gross errors (of the type you get in video) less accurate in order to get less perceivable errors (but needed to ensure color surfaces are reproduced exactly) more accurate - so most certainly the different dE is a different tolerance curve - to the point CIELAB it is not even recommended for use over 5dE.

More importantly do the test yourself - does the various dE measures match what your own eyes are telling you if used as a percent? I predicted SD vs HD would be at worst ~10% based on my own eyes and only one dE standard agrees with that. Can you perceive the Red SD as being more off HD than the blue SD or vice versa (not subjective annoyance - just pure perceptual difference). I think even yourself said the CIELAB on the JVC RS1 Green seemed out of wack with perception.

After psychovisually testing CIELUV with customers (at shootouts and calibrations - specifically L*u*v* and LCH - not L*u'v') for years now to see if the charts agree with what they are seeing (vs. CIE xy charts which they never could see how it tracks - and neither could I which is why I switched) - I am convinced that CIELUV having dE be a stronger function of Y is the reason it works so well to model light color perception severly impacted by luminance - very important in video when RGB is intentionally unbalanced in brightness.

Your NIST source simply said CIE1976 standardized on both as more perceptually more uniform than 1964 or 1931 versions. It did not say which of CIE1976 was preferred over the other for which application.
Reply
Reply
post #344 of 1810
10/20/07 at 2:25pm
Quote:
Originally Posted by TomHuffman View Post

Thus, I think that a better method for setting Contrast is to just set it at a level consistent with good color performance and reasonable light output for a given display device. What's a reasonable level?
  • CRT tubes: 35 fL
  • Plasma: 35 fL
  • LCD flat panel: 50 fL
  • Digital rear projection: 35 fL
  • Digital front projection: 15 fL
If you use HCFR, you can determine fL by multiplying the peak output (100% Y) by 0.29.

Quote:
If you are getting less than 10 fL at peak output (remember, you need to take screen gain into account), then you need a new bulb.

I have an eyeone LT probe and HCFR. Is the proper way to measure light output to:
Taking readings off of the screen without the light diffusor, then face the PJ, put on the diffusor and adjust my readings in the advanced menu and multipy by 0.29?
Reply
Reply
post #345 of 1810
10/20/07 at 3:48pm
Quote:
Originally Posted by angryht View Post

I have an eyeone LT probe and HCFR. Is the proper way to measure light output to:
Taking readings off of the screen without the light diffusor, then face the PJ, put on the diffusor and adjust my readings in the advanced menu and multipy by 0.29?

Take your 100% direct screen reading and multiply by 0.29, or to convert Tom's recommendation in SI units (cd/m^2(nits)), you'd should see a minimum of 10fL x 3.426 nits/fL = 34.26 nits for 100% stim in HCFR.
Reply
Reply
post #346 of 1810
10/20/07 at 4:46pm
Wow! If that's the case I'm at about 3 or 4 fL. Must be time for a new bulb! Thanks.
Reply
Reply
post #347 of 1810
10/22/07 at 10:44am
I see at the beginning of the thread that light meters are optional for calibration. Can light meters be used for all Hdtvs. If it can, how do convert the readings to FTL?
Reply
Reply
post #348 of 1810
10/22/07 at 10:53am
Thread Starter 
Quote:
Originally Posted by ragingd View Post

I see at the beginning of the thread that light meters are optional for calibration. Can light meters be used for all Hdtvs. If it can, how do convert the readings to FTL?

Yes.

Lux/10.76*screen gain=fL.
Reply
Reply
post #349 of 1810
10/22/07 at 1:26pm
Thanks Tom. One more question how do you determine the screen gain? I have a 50inch sxrd. Thanks for the help.
Reply
Reply
post #350 of 1810
10/22/07 at 1:36pm
TomHuffman,
Is your test dvd coded with BT.601 SMPTE-C?
Is it possible for you to add a full white screen to the test pattern RGBYCM and make it a WRGBYCM?

Thanks.
Reply
Reply
post #351 of 1810
10/22/07 at 1:46pm
Thread Starter 
Quote:
Originally Posted by ragingd View Post

Thanks Tom. One more question how do you determine the screen gain? I have a 50inch sxrd. Thanks for the help.

Screen gain is for front projection only. For rear PJ, just use direct contact with the screen and read the Lux or HCFR's Y directly
Reply
Reply
post #352 of 1810
10/22/07 at 1:55pm
Thread Starter 
Quote:
Originally Posted by yesgrey3 View Post

Is your test dvd coded with BT.601 SMPTE-C?
Is it possible for you to add a full white screen to the test pattern RGBYCM and make it a WRGBYCM?

Yes. Done.
Reply
Reply
post #353 of 1810
10/22/07 at 3:54pm
Quote:
Originally Posted by TomHuffman View Post

Yes. Done.

Thanks!
Now I can verify if the sum or RGB luminances is the same as the white's.
Reply
Reply
post #354 of 1810
10/23/07 at 6:54am
I have noticed that the two of the displays that I am trying to calibrate have an interesting characteristic in the user menu color control. Not only does the luminance, Y, value decrease for red, when I turn down the color, but the x, and y values change also. When I turn down the color the point moves towards white (less saturated). Is this typical? I thought primaries usually stay put?
Reply
Reply
post #355 of 1810
10/23/07 at 7:14am
Quote:
Originally Posted by angryht View Post

I have noticed that the two of the displays that I am trying to calibrate have an interesting characteristic in the user menu color control. Not only does the luminance, Y, value decrease for red, when I turn down the color, but the x, and y values change also. When I turn down the color the point moves towards white (less saturated). Is this typical? I thought primaries usually stay put?

This is for sure an issue with the Mits**833 series..I set the x,y with PefectTint then do the PefectColor adjustment - my points are moved by Perfect Color - no matter what - so at that point I sacrifice because there is nothing else I can do..Either have a serious oversaturated Red or an less so undersaturated Red..the others move as well - including the Secondaries - which bothers me more..
Reply
Reply
post #356 of 1810
10/23/07 at 10:14am
Thread Starter 
Quote:
Originally Posted by angryht View Post

I have noticed that the two of the displays that I am trying to calibrate have an interesting characteristic in the user menu color control. Not only does the luminance, Y, value decrease for red, when I turn down the color, but the x, and y values change also. When I turn down the color the point moves towards white (less saturated). Is this typical? I thought primaries usually stay put?

Yes. As I said in the original tutorial, these controls interact. Although the primary effect of most user Color controls is to adjust color luminance it also has a secondary effect of adjusting saturation as well.

With a properly designed CMS, the interactive effects between Saturation, Hue, and Lightness controls should be much less.
Reply
Reply
post #357 of 1810
10/23/07 at 2:38pm
Tom,

I'm trying to use your spreadsheet (delta-eCIELAB) but on 1994 sheet, the delta-e data is not updating after inputting my data in RED sections. Anything I'm missing?
Reply
Reply
post #358 of 1810
10/23/07 at 7:28pm
Thread Starter 
Quote:
Originally Posted by HDholic View Post

Tom,

I'm trying to use your spreadsheet (delta-eCIELAB) but on 1994 sheet, the delta-e data is not updating after inputting my data in RED sections. Anything I'm missing?

I just tested the CIELAB 1994 sheet available

http://www.avsforum.com/avs-vb/showp...&postcount=314

It works fine. Must be something on your end.
Reply
Reply
post #359 of 1810
10/24/07 at 3:46am
Quote:
Originally Posted by HDholic View Post

Tom,

I'm trying to use your spreadsheet (delta-eCIELAB) but on 1994 sheet, the delta-e data is not updating after inputting my data in RED sections. Anything I'm missing?


I think you had one of those old versions posted previously..I think I had the same problem as well
Reply
Reply
post #360 of 1810
10/24/07 at 6:53am
Hello Tom,

I am using an Eye1 with HFCR for calibrate the Sharp Z12K utilizing your guide for Brightness and Contrast Setting and have the following questions:

1. Is the Eye1 is sensitive enough to measure the 10IRE for brightness setting (0.63% of Y at 100%IRE). If not then the only option is to look at the moving bar for brightness setting. Correct?

2. For contrast setting, I got around 4fl with High Contrast Setting (IRIS). I could not increased the value of Y at 100IRE (increase fl) even at Max Contrast Setting. So the only option is to change the Z12k setting to High Brightness but at the expense of Contrast. Thus, a new bulb is needed only at aroung 500 Hrs. Is that the only option that I have?

3. Do I adjust brightness and contrast using the Z12K menu of I should set Brightness and Contrast at 0 setting and adjust these two parameters via setting Red Gain and Green Offset methodology.

Many thanks,
Linh
Reply
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Display Calibration

AVS Top Picks

AVS › AVS Forum › Display Devices › Display Calibration › Basic Guide to Color Calibration using a CMS (updated and enhanced)