Huge dE94 differences between ChromaPure and CalMAN - Page 2 - AVS Forum | Home Theater Discussions And Reviews
Forum Jump: 
 2Likes
Reply
 
Thread Tools
post #31 of 47 Old 03-05-2014, 02:49 PM - Thread Starter
AVS Forum Special Member
 
Chad B's Avatar
 
Join Date: Oct 2002
Location: Bradford, OH
Posts: 2,724
Mentioned: 107 Post(s)
Tagged: 0 Thread(s)
Quoted: 698 Post(s)
Liked: 1592
In CalMAN, we can specify whether to include luminance error, but not luminance weighting. In other words, if at 10% we have a significant color and gamma/level error, we can specify whether we want the reported dE to include the luminance error from gamma target in with the color or just give the color dE. But with CalMAN's undefeatable weighting, the error at 10% is going to be reported as very small whether or not luminance is included.

BTW, please bear with me in my continued use of the term "luminance weighting." I believe it is more descriptive of what CalMAN is doing than saying "absolute Y with gamma selection" or something like that...

The whole premise of the luminance weighting is very suspect in my opinion. I believe it's argued that errors at the low end are much less visually serious than errors at the bright end, but my personal experiences have not backed that up.

Let's say you have a mixed image with a high APL. Let's say that a dark object, or shadow detail, has a significant color error. Will your eyes be drawn to that color error with so many bright objects in the image close by? Maybe not. That's where I could see the premise of luminance weighting, though not to the severe degree I see here.

But what if that same dark object, with the same color error, were in a low APL picture? What if it were the only thing in the picture? Will you notice the color errors? You bet! That's why I do not believe low end errors should be pushed down into insignificance.

When I was first getting into HT and video calibration, I had a Mits CRT RPTV that had a severe positive green white balance at the low end, but things evened out somewhat at the high end. I would get so annoyed when I watched X-Files and Fox Moulder's beard stubble had a greenish tinge. But with the way this luminance weighting is, I bet it would have been reported as an acceptable dE merely because it was at the dark end of the spectrum.
PlasmaPZ80U and ConnecTEDDD like this.

touring ISF/THX calibrator with Jeti 1211 and Klein K-10
Latest reviews:JVC RS540, JVC RS3000, Sony VPL-VW695ES
Copied settings measured, Review archive
Chad B is online now  
Sponsored Links
Advertisement
 
post #32 of 47 Old 03-05-2014, 03:04 PM
AVS Forum Special Member
 
ConnecTEDDD's Avatar
 
Join Date: Sep 2010
Location: Athens, Greece
Posts: 8,781
Mentioned: 219 Post(s)
Tagged: 1 Thread(s)
Quoted: 3471 Post(s)
Liked: 4303
It will be a good idea to be able to enable/disable throu CalMAN 5.3x Settings Panel that low end perceptual chromaticity weighting.
Chad B and PlasmaPZ80U like this.

Ted's LightSpace CMS Calibration Disk Free Version for Free Calibration Software: LightSpace DPS / CalMAN ColorChecker / HCFR
S/W: LightSpace CMS, SpaceMan ICC, SpaceMatch DCM, CalMAN 5, CalMAN RGB, ChromaPure, ControlCAL
V/P: eeColor 3D LUT Box - P/G: DVDO AVLab TPG
Meters: JETI Specbos 1211, Klein K-10A, i1PRO2, i1PRO, SpectraCAL C6, i1D3, C5
ConnecTEDDD is offline  
post #33 of 47 Old 03-05-2014, 05:22 PM
AVS Forum Special Member
 
sotti's Avatar
 
Join Date: Aug 2004
Location: Seattle, WA
Posts: 7,044
Mentioned: 7 Post(s)
Tagged: 1 Thread(s)
Quoted: 339 Post(s)
Liked: 342
Quote:
Originally Posted by Chad B View Post

The whole premise of the luminance weighting is very suspect in my opinion. I believe it's argued that errors at the low end are much less visually serious than errors at the bright end, but my personal experiences have not backed that up.

Weather you agree with the premise or not, we are simply reporting the numbers the formula kicks out. We've tried to expose as many knobs as possible. But as Doug was mentioning above you can only put the data in and get the data back out.

Quote:
Originally Posted by Chad B View Post

Let's say you have a mixed image with a high APL. Let's say that a dark object, or shadow detail, has a significant color error. Will your eyes be drawn to that color error with so many bright objects in the image close by? Maybe not. That's where I could see the premise of luminance weighting, though not to the severe degree I see here.

But what if that same dark object, with the same color error, were in a low APL picture? What if it were the only thing in the picture? Will you notice the color errors? You bet! That's why I do not believe low end errors should be pushed down into insignificance.

This is a topic that we frequently discuss in the office. The best answer we have for this right now is to use something like dE uv that doesn't have any perceptive weighting in the formula in regards if that's your goal. We know we can demonstrate this concept easily with our color comparator. Toggle between the dark background and the light background and watch the lowend discoloration disappear. That's why I almost prefer the color comparator to the hard numbers, a visual representation of how far you're off can be better than anything else.
sotti is offline  
Sponsored Links
Advertisement
 
post #34 of 47 Old 03-06-2014, 02:33 AM
AVS Forum Special Member
 
TomHuffman's Avatar
 
Join Date: Jan 2003
Location: Springfield, MO
Posts: 7,911
Mentioned: 24 Post(s)
Tagged: 0 Thread(s)
Quoted: 1010 Post(s)
Liked: 842
Let's be real specific about what we are talking about.

Assume a 2.2 gamma, a Rec. 709 gamut, and a white balance of x0.319, y0.329.

Let's also assume no gamma errors.

Using CIELUV as the color difference method, if you include actual luminance values in the calculation, the dE value for this shade of gray will vary as the video level varies.

100% 5.8
90% 5.3
80% 4.8
70% 4.3
60% 3.7
50% 3.1
40% 2.5
30% 1.9
20% 1.1
10% 0.3

By any standard I am aware of, the values from 70% up indicate unacceptable white balance error. What I think is a reasonable standard indicates that 30% and below fall within tolerances. In fact, at 10% the color difference is close to unmeasurable. Between 30% and 70% is a jump ball, depending upon how demanding one's tolerances are.

But here's the thing, no competent professional would EVER judge a white balance error of x0.319, y0.329 at ANY video level between 10%-100% as acceptable. It is much too red, and easily visibly so. So the question is, what possible value is there to employing dE in a way that suggests otherwise? I really don't get it. Since x0.3127, y0.329 is the target, why not use dE in a way that reflects this? Doing so means that x0.319, y0.329 has a dE of 5.8 regardless of the video level being measured. What's the argument against this?

Tom Huffman
ChromaPure Software/AccuPel Video Signal Generators
ISF/THX Calibrations
Springfield, MO

TomHuffman is offline  
post #35 of 47 Old 03-06-2014, 04:30 AM
AVS Forum Special Member
 
Iron Mike's Avatar
 
Join Date: Oct 2006
Posts: 1,016
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 142 Post(s)
Liked: 474
Quote:
Originally Posted by TomHuffman View Post

Let's be real specific about what we are talking about.

Assume a 2.2 gamma, a Rec. 709 gamut, and a white balance of x0.319, y0.329.

Let's also assume no gamma errors.

Using CIELUV as the color difference method, if you include actual luminance values in the calculation, the dE value for this shade of gray will vary as the video level varies.

100% 5.8
90% 5.3
80% 4.8
70% 4.3
60% 3.7
50% 3.1
40% 2.5
30% 1.9
20% 1.1
10% 0.3

By any standard I am aware of, the values from 70% up indicate unacceptable white balance error. What I think is a reasonable standard indicates that 30% and below fall within tolerances. In fact, at 10% the color difference is close to unmeasurable. Between 30% and 70% is a jump ball, depending upon how demanding one's tolerances are.

But here's the thing, no competent professional would EVER judge a white balance error of x0.319, y0.329 at ANY video level between 10%-100% as acceptable. It is much too red, and easily visibly so. So the question is, what possible value is there to employing dE in a way that suggests otherwise? I really don't get it. Since x0.3127, y0.329 is the target, why not use dE in a way that reflects this? Doing so means that x0.319, y0.329 has a dE of 5.8 regardless of the video level being measured. What's the argument against this?

I generally agree with your point but the problem in your example is that you assumed "no gamma errors"...

which will rarely be the case in real world....

calibration & profiling solutions: Lightspace, Spaceman, Calman, Argyll, ColorNavigator, basICColor
profiling & calibration workflow tools: Display Calibration Tools
meter: CR-250, Klein K-10 A, i1Pro, i1D3
AVS thread: Lightspace & Custom Color Patch Set & Gamma Calibration on Panasonic 65VT60
Iron Mike is offline  
post #36 of 47 Old 03-06-2014, 06:52 AM - Thread Starter
AVS Forum Special Member
 
Chad B's Avatar
 
Join Date: Oct 2002
Location: Bradford, OH
Posts: 2,724
Mentioned: 107 Post(s)
Tagged: 0 Thread(s)
Quoted: 698 Post(s)
Liked: 1592
Quote:
Originally Posted by sotti View Post

We know we can demonstrate this concept easily with our color comparator. Toggle between the dark background and the light background and watch the lowend discoloration disappear.
I get it. Any possibility of making the background approximate real content? Not necessarily a flat 22%-25% gray; maybe thin strips of gradient ramps or something? Also I've had issues with scaling the comparator for reports, etc.

touring ISF/THX calibrator with Jeti 1211 and Klein K-10
Latest reviews:JVC RS540, JVC RS3000, Sony VPL-VW695ES
Copied settings measured, Review archive
Chad B is online now  
post #37 of 47 Old 03-06-2014, 06:57 AM - Thread Starter
AVS Forum Special Member
 
Chad B's Avatar
 
Join Date: Oct 2002
Location: Bradford, OH
Posts: 2,724
Mentioned: 107 Post(s)
Tagged: 0 Thread(s)
Quoted: 698 Post(s)
Liked: 1592
Quote:
Originally Posted by Iron Mike View Post

I generally agree with your point but the problem in your example is that you assumed "no gamma errors"...

which will rarely be the case in real world....
That wouldn't make too much difference; the perceptual luminance weighting would still probably push the results into insignificance at the low end assuming typical display gamma errors.

touring ISF/THX calibrator with Jeti 1211 and Klein K-10
Latest reviews:JVC RS540, JVC RS3000, Sony VPL-VW695ES
Copied settings measured, Review archive
Chad B is online now  
post #38 of 47 Old 03-06-2014, 09:45 AM
AVS Forum Special Member
 
sotti's Avatar
 
Join Date: Aug 2004
Location: Seattle, WA
Posts: 7,044
Mentioned: 7 Post(s)
Tagged: 1 Thread(s)
Quoted: 339 Post(s)
Liked: 342
Here's what our color comparator looks like from toms example above. you can run our triplet numbers through Lindbloom to verify that what we are showing you here maps to how the math works.


Worst case for seeing errors:



Best case for seeing errors:



Detla E datagrid (with correct 94 formula):



Without a doubt low end grayscale errors can be more visible if the brightest thing on screen is only 20-30% and you're watching in a darkened room.

But clearly just using dEuv is just as much a swing the other way in overacting. I don't have a solution, I don't do research at a university, but clearly the truth lies somewhere inbetween dE2000 and dEuv.
Chad B likes this.
sotti is offline  
post #39 of 47 Old 03-06-2014, 10:14 AM
AVS Forum Special Member
 
ConnecTEDDD's Avatar
 
Join Date: Sep 2010
Location: Athens, Greece
Posts: 8,781
Mentioned: 219 Post(s)
Tagged: 1 Thread(s)
Quoted: 3471 Post(s)
Liked: 4303
Quote:
Originally Posted by sotti View Post

Here's what our color comparator looks like from toms example above. you can run our triplet numbers through Lindbloom to verify that what we are showing you here maps to how the math works.

Lumagen or eeColor are storing the correction tables in 10-bit space, this is something your 8-bit color comparitor can't show.

For CalMAN's AutoCAL where it's generating 8-bit patterns and trying different 8-bit RGB Triplet Mixing during the Virtual Cube re-measuring multiple times each color point until it finds the accepted combination, there the Color Comparitor maybe it's enough but in real world sometimes charts/comparitors are not enough.

Also what is happening when someone runs a 31x9 with CalMAN (31-Point Luminance of WRGBCMY + 9-Point Cube) as you suggesting for eeColor users, when the low end can fixed by that extended number of measured points of low end but the software low weighting is avoiding to fix these problems?

For example if you have Klein K-10A, why not be able to fix these errors if you choose de2000 as a target dE formula for AutoCAL?

When I moved from id3 to Klein I saw amazing difference in low end to my 3D LUT using LightSpace but after using CalMAN to report the generated LUT's from LightSpace, CalMAN has reporting the same dE at my low end.... That's why is needed to add a button for dE2000 to enable or disable this perceptual filtering of dE of low end.

LightSpace or ChromaPure or HCFR are measuring according to dE standarts, so no dE hidding of errors.... Is there any published document where it says that dE numbers at low end is not needed or we can't see them?
prsut likes this.

Ted's LightSpace CMS Calibration Disk Free Version for Free Calibration Software: LightSpace DPS / CalMAN ColorChecker / HCFR
S/W: LightSpace CMS, SpaceMan ICC, SpaceMatch DCM, CalMAN 5, CalMAN RGB, ChromaPure, ControlCAL
V/P: eeColor 3D LUT Box - P/G: DVDO AVLab TPG
Meters: JETI Specbos 1211, Klein K-10A, i1PRO2, i1PRO, SpectraCAL C6, i1D3, C5
ConnecTEDDD is offline  
post #40 of 47 Old 03-06-2014, 10:27 AM
AVS Forum Special Member
 
sotti's Avatar
 
Join Date: Aug 2004
Location: Seattle, WA
Posts: 7,044
Mentioned: 7 Post(s)
Tagged: 1 Thread(s)
Quoted: 339 Post(s)
Liked: 342
Quote:
Originally Posted by ConnecTEDDD View Post

Lumagen or eeColor are storing the correction tables in 10-bit space, this is something your 8-bit color comparitor can't show.

The color comparator is only 8-bit resolution yes, but that's independent of calibration. This is just a reporting tool.

Quote:
Originally Posted by ConnecTEDDD View Post

Also what is happening when someone runs a 31x9 with CalMAN (31-Point Luminance of WRGBCMY + 9-Point Cube) as you suggesting for eeColor users, when the low end can fixed by that extended number of measured points of low end but the software low weighting is avoiding to fix these problems?
You can choose the dE formula you want and the dE targets you want. If you don't like the fromula we choose by default change it. In our experience we've selected the best defaults.
Quote:
Originally Posted by ConnecTEDDD View Post

For example if you have Klein K-10A, why not be able to fix these errors if you choose de2000 as a target dE formula for AutoCAL?

Sure lower the targets, change the formula, we have to ship targets that are broadly compatible.
Quote:
Originally Posted by ConnecTEDDD View Post

LightSpace or ChromaPure or HCFR are measuring according to dE standarts, so no dE hidding of errors.... Is there any published document where it says that dE numbers at low end is not needed or we can't see them?

By default we don't hide anything. We have all of the above formulas, all of them take the X,Y,Z data normalized 0-1 from the 100% reading and run them through the formulas. The options to tweak dE reporting are available for any user to change on chart or place we report it in the app. 99% of the time we are using raw dE 2000.
sotti is offline  
post #41 of 47 Old 03-06-2014, 11:05 AM
AVS Forum Special Member
 
sotti's Avatar
 
Join Date: Aug 2004
Location: Seattle, WA
Posts: 7,044
Mentioned: 7 Post(s)
Tagged: 1 Thread(s)
Quoted: 339 Post(s)
Liked: 342
Looking at only the 10% swatch the triplets are 19,19,19 for the colorchecker reference swatch (calculated on an sRGB gamma curve for a power 2.2 10% target). and 20,19,19 for the measured sample. In Tom's example the observed color is +0.0063 x, 0 x, and 0 Y from the target


On my display
Reference measured x: 0.3393 y: 0.3584 Yn: 0.0040
Observed measured x: 0.3461 y: 0.3578 Yn: 0.0042
difference x: 0.0068 y: 0.0006 Yn: 0.0002


So at least on this HP LP2465 monitor the color comparator gives extremely close results to our theoretical test is. No one can honestly look at those two colors and say that is a dE of almost 6.

Using the above x,y,Yn numbers converting to Lab and getting the dE's
dE76: 0.4, dE94: 0.39, dE2000 0.42
sotti is offline  
post #42 of 47 Old 03-06-2014, 12:01 PM
Senior Member
 
prsut's Avatar
 
Join Date: Sep 2007
Location: Slovakia
Posts: 261
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 17 Post(s)
Liked: 65
Quote:
Originally Posted by ConnecTEDDD View Post


When I moved from id3 to Klein I saw amazing difference in low end to my 3D LUT using LightSpace but after using CalMAN to report the generated LUT's from LightSpace, CalMAN has reporting the same dE at my low end.... That's why is needed to add a button for dE2000 to enable or disable this perceptual filtering of dE of low end.

LightSpace or ChromaPure or HCFR are measuring according to dE standarts, so no dE hidding of errors.... Is there any published document where it says that dE numbers at low end is not needed or we can't see them?

I do not have Klein, I am no expert, but I am +1 with Ted. One simple reason for every noob, self made, home calibrator like me.
Doing 2pt WB at 20-100, setting low is affecting high and vice versa, BUT, how I know that high IS affecting low if dE2000 (CalMAN recommended) is better than good on low ???. On my GT50 plasma I have low end set in SM, very precise, to not have any touch of rgb push. I am avoiding to move rgb low in user menu then. But what if rgb gain is affecting my perfectly set low ? I know from my sessions, gain IS affecting low on my GT50, but till today I ignored it because of dE2000 at 20%IRE is about 0.5.
ConnecTEDDD likes this.

LG 55EG910V, Oppo BD-103, miniDSP nanoAVR DL, Yamaha RX-V685, DALI Oberon3 + Oberon Vokal + Spektor 1, XTZ SUB 10.17,
prsut is offline  
post #43 of 47 Old 03-06-2014, 12:23 PM
AVS Forum Special Member
 
sotti's Avatar
 
Join Date: Aug 2004
Location: Seattle, WA
Posts: 7,044
Mentioned: 7 Post(s)
Tagged: 1 Thread(s)
Quoted: 339 Post(s)
Liked: 342
Quote:
Originally Posted by prsut View Post

I do not have Klein, I am no expert, but I am +1 with Ted. One simple reason for every noob, self made, home calibrator like me.
Doing 2pt WB at 20-100, setting low is affecting high and vice versa, BUT, how I know that high IS affecting low if dE2000 (CalMAN recommended) is better than good on low ???. On my GT50 plasma I have low end set in SM, very precise, to not have any touch of rgb push. I am avoiding to move rgb low in user menu then. But what if rgb gain is affecting my perfectly set low ? I know from my sessions, gain IS affecting low on my GT50, but till today I ignored it because of dE2000 at 20%IRE is about 0.5.

This is what's been coming up a bit in the 3D LUT thread. the 20% value is likely 0.5 from target, but when you look at a grayscale ramp you are taking in A LOT more data than a single measurement.

In order to detect the red push you'd likely need to measure every 4 bits, if you do that you'll likely see errors pop up at some steps that are into the 2's.

Let me ask you directly, what dE do you see there for bit level 26? I measured it on my display an the numbers came out similarly to describe the kinds of errors we are talking about. We need to isolate our variables, if we are only talking about the how well dE describes the difference between to colors then lets focus on those two RGB swatches that seem to represent exactly the subject at hand.

I assert that 0.5 dE is about right for the difference between those two colors, maybe it could be higher, but no worse than 1, being that 1 is intended as the minimum visible difference.


A different topic is how do we use dE to quantify grayscale performance, and i have some ideas on that. Clearly taking measurements every 10% is not sufficient to communicate what is actually happening.
sotti is offline  
post #44 of 47 Old 03-06-2014, 03:40 PM
AVS Forum Special Member
 
fhoech's Avatar
 
Join Date: Apr 2013
Location: Stuttgart, Germany
Posts: 1,056
Mentioned: 73 Post(s)
Tagged: 0 Thread(s)
Quoted: 579 Post(s)
Liked: 380
What might be an idea for an evaluation control is an APL slider, which could be used to scale the measured values and thus affect the way the various dE formulas react to the input values. An example:

Say you have the usual grayscale measurements, n% steps. Now you set the APL slider to 30%, which would scale the 0%..30% grayscale measurements into the 0..1 (0..100 L*) range (obviously clipping/filtering out anything above 30%). This has the effect that dE values for dark colors would increase accordingly, maybe providing a better correspondence to what can be visually observed in dark movie scenes.

DisplayCAL - Display Calibration and Characterization powered by ArgyllCMS
Current stable version 3.8.9.1 released 2019-12-07 | Changelog
DisplayCAL on Facebook
fhoech is offline  
post #45 of 47 Old 03-07-2014, 01:23 AM
AVS Forum Special Member
 
HDTVChallenged's Avatar
 
Join Date: Jan 2002
Posts: 8,864
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 3 Post(s)
Liked: 222
Quote:
Originally Posted by Iron Mike View Post

I generally agree with your point but the problem in your example is that you assumed "no gamma errors"...

which will rarely be the case in real world....

Which brings up a mea culpa. Earlier in the thread, I talked about the option to include Luminance in the dE calculation. The specific software option I was talking about was actually labeled 'Use target gamma in dE calculations.'

Same meaning/context ... AFAIK ... dE on (x,y) alone vs dE on (x,y,Y).

Just a clarification in case anyone got hung up on my earlier terminology. smile.gif
HDTVChallenged is offline  
post #46 of 47 Old 03-12-2014, 02:06 PM
AVS Forum Special Member
 
ConnecTEDDD's Avatar
 
Join Date: Sep 2010
Location: Athens, Greece
Posts: 8,781
Mentioned: 219 Post(s)
Tagged: 1 Thread(s)
Quoted: 3471 Post(s)
Liked: 4303
Quote:
Originally Posted by sotti View Post

Quote:
Originally Posted by prsut View Post

In what version is it fixed ?

5.3.0.1440 or better.

Hi Joel, is the 5.3.0.1450 release fixing the dE calculation error?

I don't see that fix in the release notes.

Ted's LightSpace CMS Calibration Disk Free Version for Free Calibration Software: LightSpace DPS / CalMAN ColorChecker / HCFR
S/W: LightSpace CMS, SpaceMan ICC, SpaceMatch DCM, CalMAN 5, CalMAN RGB, ChromaPure, ControlCAL
V/P: eeColor 3D LUT Box - P/G: DVDO AVLab TPG
Meters: JETI Specbos 1211, Klein K-10A, i1PRO2, i1PRO, SpectraCAL C6, i1D3, C5
ConnecTEDDD is offline  
post #47 of 47 Old 03-12-2014, 02:31 PM
AVS Forum Special Member
 
sotti's Avatar
 
Join Date: Aug 2004
Location: Seattle, WA
Posts: 7,044
Mentioned: 7 Post(s)
Tagged: 1 Thread(s)
Quoted: 339 Post(s)
Liked: 342
Quote:
Originally Posted by ConnecTEDDD View Post

Hi Joel, is the 5.3.0.1450 release fixing the dE calculation error?

I don't see that fix in the release notes.

Yes it is, I'll see if we can't get that into the notes.
ConnecTEDDD likes this.
sotti is offline  
Sponsored Links
Advertisement
 
Reply Display Calibration

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