HCFR - Open source projector and display calibration software - Page 141 - AVS Forum
Forum Jump: 
 53Likes
Reply
 
Thread Tools
post #4201 of 4225 Old 08-26-2014, 09:35 PM
AVS Special Member
 
HDTVChallenged's Avatar
 
Join Date: Jan 2002
Posts: 8,353
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 3 Post(s)
Liked: 122
Quote:
Originally Posted by xvfx View Post
I'm surprised most don't use CIE2000.
dE76 is the most ruthless tool for greyscale. ... IMO, one should remove as much absolute mathematical error as possible regardless of so-called "perceptual" weighing.

CIE2000 is for when you want to "feel good" about the errors you can't fix.
HDTVChallenged is offline  
Sponsored Links
Advertisement
 
post #4202 of 4225 Old 08-26-2014, 10:30 PM
Advanced Member
 
gwgill's Avatar
 
Join Date: Jan 2013
Location: Melbourne, Australia
Posts: 588
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 45 Post(s)
Liked: 75
Quote:
Originally Posted by HDTVChallenged View Post
dE76 is the most ruthless tool for greyscale. ... IMO, one should remove as much absolute mathematical error as possible regardless of so-called "perceptual" weighing
There isn't a whole lot of difference between DE76, DE94 or CIE2000 near the neutral axis - most of the tweaks are in the saturated colors which tend to be de-weighted in DE94 and CIE2000.
gwgill is offline  
post #4203 of 4225 Old 08-27-2014, 08:52 AM
AVS Special Member
 
HDTVChallenged's Avatar
 
Join Date: Jan 2002
Posts: 8,353
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 3 Post(s)
Liked: 122
Quote:
Originally Posted by gwgill View Post
There isn't a whole lot of difference between DE76, DE94 or CIE2000 near the neutral axis - most of the tweaks are in the saturated colors which tend to be de-weighted in DE94 and CIE2000.
All I'm saying is that little "errors" add up in a Y/Cr/Cb color system ... if your "zero" is really something like "0.1," that just makes it harder to find all the other colors. So be as ruthless as possible when you can.
HDTVChallenged is offline  
post #4204 of 4225 Old 08-27-2014, 07:24 PM - Thread Starter
AVS Special Member
 
zoyd's Avatar
 
Join Date: Sep 2006
Location: Planet Dog
Posts: 4,449
Mentioned: 10 Post(s)
Tagged: 0 Thread(s)
Quoted: 91 Post(s)
Liked: 313
HCFR - ArgyllCMS meter correction comparison

I wanted to compare the meter correction technique that HCFR uses (RGBW 4-color matrix solution) to what ArgyllCMS is doing to evaluate any differences. The data below was obtained via the following:

1. Profile a D3 against the JETI-1211 using ArgyllCMS
2. Measure a test set with the JETI-1211 and the profiled D3
3. Profile a D3 against the JETI-1211 using HCFR
4. Measure the test set again with the HCFR profiled D3
5. Remeasure with the JETI-1211 for repeatability

In Step 1 ArgyllCMS calculates the best-fit matrix that minimizes color difference errors between the two probes (as opposed to the closed form solution of the 4-color matrix method). For this method I used a profiling set of 22 colors consisting of 25% increments along each color axis and ~11% increments along the neutral axis.

Steps 2-5 use a different set of 37 test colors at higher density along each RGBW axis.

Steps 3 and 4 were repeated using both 100% RGBW and 75% RGBW patches to form the HCFR correction matrix.

Results show that the ArgyllCMS method reproduces the reference values somewhat better than the 4-color method and the 75% level-based a tad better than the 100%, although the absolute color differences are pretty small (mean dE2000: best-fit =0.19, 75% 4-color=0.26, and 100% 4-color=0.39). Repeatability was excellent at 0.09 dE2000 mean difference.

gwgill and spacediver like this.
zoyd is online now  
post #4205 of 4225 Old 08-27-2014, 08:58 PM
Member
 
James Freeman's Avatar
 
Join Date: Sep 2013
Posts: 151
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 65 Post(s)
Liked: 25
Can someone please explain why do we need a spectrophotometer along with a colorimiter?
Are the results a lot better when using a correction matrix?

I've been calibrating my displays with i1D2 then with i1D3 for years, was is "not accurate"?
Is a difference of 0.5 THAT visible?
James Freeman is offline  
post #4206 of 4225 Old 08-27-2014, 09:05 PM
Advanced Member
 
gwgill's Avatar
 
Join Date: Jan 2013
Location: Melbourne, Australia
Posts: 588
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 45 Post(s)
Liked: 75
Quote:
Originally Posted by zoyd View Post
Results show that the ArgyllCMS method reproduces the reference values somewhat better than the 4-color method and the 75% level-based a tad better than the 100%, although the absolute color differences are pretty small (mean dE2000: best-fit =0.19, 75% 4-color=0.26, and 100% 4-color=0.39). Repeatability was excellent at 0.09 dE2000 mean difference.
That's pretty interesting, since I haven't done such a comparison myself.

Any feeling whether Argyll's fitting makes any difference (better or worse) when just using 4 patches (RGBW) ?
gwgill is offline  
post #4207 of 4225 Old 08-27-2014, 09:06 PM
Advanced Member
 
gwgill's Avatar
 
Join Date: Jan 2013
Location: Melbourne, Australia
Posts: 588
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 45 Post(s)
Liked: 75
Quote:
Originally Posted by James Freeman View Post
Can someone please explain why do we need a spectrophotometer along with a colorimiter?
Are the results a lot better when using a correction matrix?
The point is to use the Spectrometer to create the correction matrix.
gwgill is offline  
post #4208 of 4225 Old 08-27-2014, 09:09 PM
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 792
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 49 Post(s)
Liked: 66
Quote:
Originally Posted by James Freeman View Post
Is a difference of 0.5 THAT visible?
if I understand correctly, the 0.5 is the difference between the i1d3 and jeti after profiling, not before.
spacediver is offline  
post #4209 of 4225 Old 08-27-2014, 09:22 PM
Member
 
James Freeman's Avatar
 
Join Date: Sep 2013
Posts: 151
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 65 Post(s)
Liked: 25
Quote:
Originally Posted by gwgill View Post
The point is to use the Spectrometer to create the correction matrix.
I understand, Thanks.
Will a spectrometer improve the calibration of the Greyscale or only the Colors?

From what I've seen I think that a well behaved Greyscale and Gamma are much more important than sub 3dE color errors.


Panasonic TC-P50ST60.
HTPC (MPC-HC, MadVR), PS3.
Studio Monitors + Polk Audio PSW110 Subwoofer.
James Freeman is offline  
post #4210 of 4225 Old 08-27-2014, 09:50 PM
Advanced Member
 
gwgill's Avatar
 
Join Date: Jan 2013
Location: Melbourne, Australia
Posts: 588
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 45 Post(s)
Liked: 75
Quote:
Originally Posted by James Freeman View Post
Will a spectrometer improve the calibration of the Greyscale or only the Colors?
There's no guarantee that using a spectrometer will improve the accuracy of your colorimeter - by accident, the colorimeter may be more accurate with one of the standard calibration matricies.

But the point of using a spectrometer is that it will guarantee a minimum level of accuracy, irrespective of how good or bad the colorimeter and it's standard calibration matricies fit your display.

Since we are more visually sensitive to tints in neutrals, any improvement in accuracy is likely to be noticed in those areas.
Brian Hampton likes this.
gwgill is offline  
post #4211 of 4225 Old 08-27-2014, 10:07 PM
AVS Special Member
 
sillysally's Avatar
 
Join Date: Jun 2006
Location: Chicago
Posts: 3,714
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 73 Post(s)
Liked: 262
Quote:
Originally Posted by gwgill View Post
That's pretty interesting, since I haven't done such a comparison myself.

Any feeling whether Argyll's fitting makes any difference (better or worse) when just using 4 patches (RGBW) ?
As you may know I have been asking the above question, not just for ArgyllCMS but also for CM and LS.
For my meter correction matrix, I have been using Klein's ChromaPure software for the K10-A and Jeti 1211.
How does ArgyllCMS "best fit" work.?
I did try and use dispcal/Argyll using the beta binary release for the K10-A, but didn't think it worked very well to create a meter correction matrix.
When I tried to run a meter correction matrix for the K10-A, dispcalGUI powered by ArgyllCMS used only the 4 WRGB patterns.
Does best fit use more than one four set of WRGB?

Because I use a VT60 (Plasma), I find that with using the Jeti 1211 as reference meter and K10-A as source meter to make a meter correction matrix. Using Klein's ChromaPure and make 3 or 4 meter correction matrix's, then use the correction matrix that has the best fit (least amount of errors for WRGB for use with the K10-A).

btw, are you planing on releasing the new source code for the K10-A in a binary release that I can use with dispcal.?
Also are you planing any time soon to add support for the Lumagen Radiance 20XX ?

ss
sillysally is offline  
post #4212 of 4225 Old 08-27-2014, 10:23 PM
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 792
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 49 Post(s)
Liked: 66
Zoyd/Graeme/anyone else,

I'm trying an interesting experiment with gratings to see if I can infer luminance levels below that of the instrument's range (i1d3). The approach is as follows:

take a grating with spatially alternating darker and lighter lines. Measure the luminance of each of these values separately, and then measure the grating (the colorimeter's sensor is large enough that it integrates the luminance information across the stripes). Preliminary calculations show that this works. However, when I created a grating with a black stripe and a lighter stripe, the lighter stripe measured 0.28 cd/m2, the grating itself measured at 0.14 cd/m2, implying the dark stripe (black) was exactly 0 cd/m2.

Is there anyway to gain access to more than 2 decimal places of readoff precision in the real time readings in HCFR? And do you think this approach could work to infer luminances below that of the instrument's range?
spacediver is offline  
post #4213 of 4225 Old 08-27-2014, 10:47 PM
Advanced Member
 
gwgill's Avatar
 
Join Date: Jan 2013
Location: Melbourne, Australia
Posts: 588
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 45 Post(s)
Liked: 75
Quote:
Originally Posted by sillysally View Post
How does ArgyllCMS "best fit" work.?
It creates a matrix to minimize the delta E's between the measured values.
Quote:
When I tried to run a meter correction matrix for the K10-A, dispcalGUI powered by ArgyllCMS used only the 4 WRGB patterns.
Does best fit use more than one four set of WRGB?
It uses best fit for however many patches are measured, including the default WRGB. You can select a larger number of patches using the "ccxxmake -s N" flag, where N > 1, and it will generate N x N x N regular spaced cube test patches.

A disadvantage of using ccxxmake for the K10 is that currently it has no capability to store the matrix in the meter.
Quote:
btw, are you planing on releasing the new source code for the K10-A in a binary release that I can use with dispcal.?
The current 1.7 beta snapshot supports the K10-A.
http://www.argyllcms.com/beta1.7_win32_exe.zip
http://www.argyllcms.com/beta1.7_win64_exe.zip
Quote:
Also are you planing any time soon to add support for the Lumagen Radiance 20XX ?
I don't have a Radiance 20XX, so that's not possible. If Lumagen want ArgyllCMS support, they know what to do...
gwgill is offline  
post #4214 of 4225 Old 08-27-2014, 10:52 PM
Advanced Member
 
gwgill's Avatar
 
Join Date: Jan 2013
Location: Melbourne, Australia
Posts: 588
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 45 Post(s)
Liked: 75
Quote:
Originally Posted by spacediver View Post
take a grating with spatially alternating darker and lighter lines. Measure the luminance of each of these values separately, and then measure the grating (the colorimeter's sensor is large enough that it integrates the luminance information across the stripes).
It's hard to be sure what you are suggesting, but assuming you are talking about printing black stripes on a transparent medium, this will reduce the light level reaching the meter, worsening it's ability to read low light levels.

If you mean something else, you'll have to make your description clearer, and/or include some photo's.
gwgill is offline  
post #4215 of 4225 Old 08-27-2014, 11:17 PM
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 792
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 49 Post(s)
Liked: 66
Quote:
Originally Posted by gwgill View Post
It's hard to be sure what you are suggesting, but assuming you are talking about printing black stripes on a transparent medium, this will reduce the light level reaching the meter, worsening it's ability to read low light levels.

If you mean something else, you'll have to make your description clearer, and/or include some photo's.
Oh this is all done on a display. I use matlab to create the test patterns, like this one:



This pattern contains alternating stripes of black (video level 0) and dark grey (video level 20). Measuring the grating with a colorimeter will integrate the luminance information coming from both sets of stripes. My video level 20 (RGB = [20 20 20]) measures at 0.28 nits, and the grating as a whole measures at 0.14 nits. This implies that video level 0 is 0 nits. If I had access to more decimal places (e.g. 0.2843 nits for video level 20, and 0.1438 for the grating), then I could infer a black level of 0.0033 nits.
spacediver is offline  
post #4216 of 4225 Old Yesterday, 04:16 AM
Advanced Member
 
gwgill's Avatar
 
Join Date: Jan 2013
Location: Melbourne, Australia
Posts: 588
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 45 Post(s)
Liked: 75
Quote:
Originally Posted by spacediver View Post
My video level 20 (RGB = [20 20 20]) measures at 0.28 nits, and the grating as a whole measures at 0.14 nits. This implies that video level 0 is 0 nits. If I had access to more decimal places (e.g. 0.2843 nits for video level 20, and 0.1438 for the grating), then I could infer a black level of 0.0033 nits.
I'm not sure you've thought that out. Diluting what you're trying to measure with something that may be an order of magnitude brighter (given the typical gamma) is no way to improve precision. If there were more decimal places in the reading, then you would be able to measure the black level directly.
gwgill is offline  
post #4217 of 4225 Old Yesterday, 06:56 AM - Thread Starter
AVS Special Member
 
zoyd's Avatar
 
Join Date: Sep 2006
Location: Planet Dog
Posts: 4,449
Mentioned: 10 Post(s)
Tagged: 0 Thread(s)
Quoted: 91 Post(s)
Liked: 313
Quote:
Originally Posted by James Freeman View Post
Can someone please explain why do we need a spectrophotometer along with a colorimiter?
Are the results a lot better when using a correction matrix?

I've been calibrating my displays with i1D2 then with i1D3 for years, was is "not accurate"?
Is a difference of 0.5 THAT visible?
No, 0.5 dE2000 is not perceptible but that was after calibrating it to the spectrometer. Without a spectrometer based correction the amount of residual error you might have will depend on the colorimeter, the display, and any vendor supplied corrections. In this case no correction yields peak errors greater than 3 and with the vendor supplied plasma correction peak errors of ~1.5 (see table below).

Quote:
Originally Posted by gwgill View Post
That's pretty interesting, since I haven't done such a comparison myself.

Any feeling whether Argyll's fitting makes any difference (better or worse) when just using 4 patches (RGBW) ?
I was not expecting much of a difference given the D3 has excellent repeatability and linearity. I repeated the experiment and added the ArgyllCMS version of 4-color correction and found the following:



A very marginal benefit when adding the extra colors.

Quote:
Originally Posted by spacediver View Post
If I had access to more decimal places (e.g. 0.2843 nits for video level 20, and 0.1438 for the grating), then I could infer a black level of 0.0033 nits.
The freemeasures tab will show 3 decimals for the Y value in continuous measure mode but I agree with Graeme, I don't see how you get more precision out of the sensor with your test.
gwgill likes this.
zoyd is online now  
post #4218 of 4225 Old Yesterday, 01:14 PM
Advanced Member
 
xvfx's Avatar
 
Join Date: Feb 2013
Location: UK
Posts: 623
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 71 Post(s)
Liked: 64
Quote:
Originally Posted by HDTVChallenged View Post
dE76 is the most ruthless tool for greyscale. ... IMO, one should remove as much absolute mathematical error as possible regardless of so-called "perceptual" weighing.

CIE2000 is for when you want to "feel good" about the errors you can't fix.
Quote:
Originally Posted by HDTVChallenged View Post
All I'm saying is that little "errors" add up in a Y/Cr/Cb color system ... if your "zero" is really something like "0.1," that just makes it harder to find all the other colors. So be as ruthless as possible when you can.
Hmm… and the rest of that thread by the big names were very interesting.

What dE to use when calibrating?
xvfx is offline  
post #4219 of 4225 Old Yesterday, 02:06 PM
Senior Member
 
<^..^>Smokey Joe's Avatar
 
Join Date: Aug 2006
Location: Christchurch, New Zealand
Posts: 377
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 12 Post(s)
Liked: 46
@zoyd

You may be more likely to see a better correction of the RGBW if you use 10%, 30%, 70% and 90% levels of RGBW.
My reasoning behind this is the spectral distribution changes as you get closer to black. Simply, the hollows can drop below detection and any peaks dominate more which could cause slight metameric failure.

You would use the 90% as the anchor and then apply a simple y=bx+c or use the regression x=(y-c)/b to each RGBW, depends which way you plot the slope and intercept.

A simple test to see if an improvement could be found is to do a standard RGBW correction at 10%, 30%, 70% and 90% levels. Plot the RGBW results and see if they drift. I suspect some displays will drift others won't, depends on the technology. Might even find different regression paths for each RGBW.
ConnecTEDDD likes this.

Masterpiece Calibration Ltd
Christchurch NZ

Last edited by <^..^>Smokey Joe; Yesterday at 02:18 PM.
<^..^>Smokey Joe is offline  
post #4220 of 4225 Old Yesterday, 03:58 PM
Advanced Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 792
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 49 Post(s)
Liked: 66
Quote:
Originally Posted by gwgill View Post
I'm not sure you've thought that out. Diluting what you're trying to measure with something that may be an order of magnitude brighter (given the typical gamma) is no way to improve precision. If there were more decimal places in the reading, then you would be able to measure the black level directly.
Take an instrument that measures with a precision of 0.01 cd/m2 (and can read down to 0.01 cd/m2).

Now take a grating that has alternating lines, with respective luminances of exactly 0.003 cd/m2 and 0.08 cd/m2. The grating will "output" a luminance of exactly 0.0415 cd/m2.

The instrument will measure the lighter line as 0.08 cd/m2, and will measure the grating as 0.04 cd/m2.

Doing the math, one can infer that the darker line = (grating luminance*2)-lighter line, = 0.08 - 0.08 = 0.

Damn, I see your point.

Ok, back to psychophysics
spacediver is offline  
post #4221 of 4225 Old Yesterday, 04:24 PM
Member
 
Join Date: Jul 2002
Location: Boston, MA
Posts: 166
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 11
Hi All, I'm looking to contribute to the development of HCFR. Is the hcfr-development list at sourceforge still a good place to ask questions? (There hasn't been a post since February) Do the devs hang out anywhere else (irc, other mailing lists, etc?)
JStanton is online now  
post #4222 of 4225 Old Yesterday, 04:54 PM - Thread Starter
AVS Special Member
 
zoyd's Avatar
 
Join Date: Sep 2006
Location: Planet Dog
Posts: 4,449
Mentioned: 10 Post(s)
Tagged: 0 Thread(s)
Quoted: 91 Post(s)
Liked: 313
Quote:
Originally Posted by <^..^>Smokey Joe View Post
@zoyd

You may be more likely to see a better correction of the RGBW if you use 10%, 30%, 70% and 90% levels of RGBW.
I have no desire to do multiple 4-color profiles to gain a perhaps 0.15 dE better match to the reference probe especially since the ArgyllCMS result is already near repeatability limits, and I can just use that.
zoyd is online now  
post #4223 of 4225 Old Yesterday, 06:16 PM - Thread Starter
AVS Special Member
 
zoyd's Avatar
 
Join Date: Sep 2006
Location: Planet Dog
Posts: 4,449
Mentioned: 10 Post(s)
Tagged: 0 Thread(s)
Quoted: 91 Post(s)
Liked: 313
Quote:
Originally Posted by JStanton View Post
Hi All, I'm looking to contribute to the development of HCFR. Is the hcfr-development list at sourceforge still a good place to ask questions? (There hasn't been a post since February) Do the devs hang out anywhere else (irc, other mailing lists, etc?)
You may want to sign up on the ArgyllCMS mailing list as well but you can reach me or Graeme on the HCFR list (maybe JohnAD as well). I'm the only one working on the code at the moment and have been too busy with other things to get much done recently.
zoyd is online now  
post #4224 of 4225 Old Yesterday, 07:22 PM
AVS Special Member
 
sillysally's Avatar
 
Join Date: Jun 2006
Location: Chicago
Posts: 3,714
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 73 Post(s)
Liked: 262
Quote:
Originally Posted by gwgill View Post
I have the snapshot from May for beta 1.7

According to dispcalGUI beta, there is a updated version beta 1.7,

Quote:
Originally Posted by gwgill View Post
I don't have a Radiance 20XX, so that's not possible. If Lumagen want ArgyllCMS support, they know what to do...
I thought Jim from Lumagen was going to send you one. At-least that is what he said he would do.

ss
sillysally is offline  
post #4225 of 4225 Old Yesterday, 07:51 PM
Advanced Member
 
gwgill's Avatar
 
Join Date: Jan 2013
Location: Melbourne, Australia
Posts: 588
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 45 Post(s)
Liked: 75
Quote:
Originally Posted by sillysally View Post
I have the snapshot from May for beta 1.7
According to dispcalGUI beta, there is a updated version beta 1.7
There is, but I wouldn't recommend it for general use at the moment since there
are some peculiarities with profiling that I'm still sorting out.
Quote:
I thought Jim from Lumagen was going to send you one. At-least that is what he said he would do.
We had a discussion and there was some suggestion that they were willing to do that, but at this point they haven't followed up on it.
gwgill is offline  
Reply Display Calibration

User Tag List

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off