Lumagen Vision 24p fixup release to include gamut control - Page 4 - AVS Forum
Forum Jump: 
Reply
 
Thread Tools
post #91 of 736 Old 05-13-2008, 08:37 AM
AVS Special Member
 
Lawguy's Avatar
 
Join Date: Sep 2004
Posts: 5,705
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Quote:
Originally Posted by TomHuffman View Post

I have made several constructive suggestions. I have suggested that if you use use the gamut control, you should use it on one primary only to minimize the effects on the secondaries. I have suggested that you not use color decoding controls to address errors in saturation. I have suggested that if you are really interested in color correction that you consider springing for the Radiance, which offers a CMS that works exactly as advertised. While we are on the subject, it would be useful to provide pre-calibration numbers (all adjustments zeroed out) so we can have a baseline against which to judge your efforts.

Tom,

If you, and others, are agreeable, I would like to start fresh. Why don't we do this in the correct way?

There are at least three of us that are trying this. The correct way to do this, is to start as you suggest. We should all post baseline measurements that represent out of the box performance. Why not also post settings with the color control at -15, but no other controls at work.

The next step is where the many paths diverge because there are several different strategies that appear to work. (1) Vision gamut controls only; (2) Vision gamut controls plus adjustments with other Vision controls (COLR, HUE, etc); (3) Vision gamut plus JVC color controls; and (4)All of the above (all vision and JVc controls).

Tom, if you are interested and the others agree, I would ask that we all work together under your direction. Tell us what to do and we will do it.

It really seems to me that this is all working. I don't know if we can get things to work any better than they are now, but it is possible.

What do you say?

Affable Nitwit
Lawguy is offline  
Sponsored Links
Advertisement
 
post #92 of 736 Old 05-13-2008, 10:55 AM
AVS Club Gold
 
mrlittlejeans's Avatar
 
Join Date: Jan 2004
Location: the area formerly known as Black Forest
Posts: 4,714
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 192
Let me add that my HDP should arrive this week and I will be able to play with it next week. I will need to buy a probe though. Any recommendations under $500? I will be happy to post all my settings as well.

Absence of evidence is not evidence of absence

mrlittlejeans is offline  
post #93 of 736 Old 05-13-2008, 12:01 PM - Thread Starter
jrp
AVS Special Member
 
jrp's Avatar
 
Join Date: Jan 2003
Location: Portland, Oregon, USA
Posts: 1,025
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 13 Post(s)
Liked: 25
We appreciate everyone's efforts to better understand the Vision series Gamut and it's limitation. I want to specifically thank Tom for his efforts, and say his comments do not seem harsh at all. Facts often come across as bunt.

Tom is right that the Vision series does not provide gamut correction. Rather, it is gamut mitigation - if you understand gamut measurements and are willing to take the time to iterate and balance between conflicting requirements.

Tom is also correct in suggesting that if you want to "correct" Gamut you should use a Radiance, since you can achieve much better results using it's linear-gamma 3D palette.

As I have previously mentioned, we debated even releasing the Vision gamut knowing it's limitation. We were trying to provide extra value in the Vision series at no additional cost to our user base. I believe we have, but am sorry that it seems to have created a storm as well. Certainly this thread has provided useful information about gamut that might not have come out otherwise. So, maybe it's a good storm.

Jim Peterson
Lumagen
jrp is offline  
post #94 of 736 Old 05-13-2008, 01:38 PM
Senior Member
 
joikd's Avatar
 
Join Date: Nov 2006
Posts: 378
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 1 Post(s)
Liked: 11
Good information here.

I have a VisionHDP and an Eye One Display LT. I'm new at this, but am willing to provide measurement data, or anything else--just let me know (and maybe tell me how).
joikd is offline  
post #95 of 736 Old 05-13-2008, 02:35 PM
AVS Special Member
 
reincarnate's Avatar
 
Join Date: Jul 2004
Posts: 2,989
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by TomHuffman View Post

I don't usually respond to your posts because I make it a policy to ignore dedicated Internet trolls, but this requires an exception.
http://www.avsforum.com/avs-vb/showt...9#post13003459

Tom,
To tell the truth, I think you largely do good here, but you and Mr. Sorel did have a personal vendetta against JVC that went on for too many months.
You wisely backed off when your repuation began to suffer and have largely taken the high road since. But this childish name calling above must be a slip of desperation.
Consider being a bit more consistent? For example why not voice objections to the $15K Marantz S2 with its widely inaccurate and over-saturated greens? It has no CMS whatsoever. Inaccurate color is inaccurate color, irrespective of the manufacture. Thank you!
reincarnate is offline  
post #96 of 736 Old 05-13-2008, 02:49 PM
AVS Special Member
 
reincarnate's Avatar
 
Join Date: Jul 2004
Posts: 2,989
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Same too goes for Lumagen. We have heard your voice clearly and its getting old. Why not go find the faults of the many other Video Processors? (Maybe you have as I don't read this forum too often). They are chocked full of bugs way worse than any Lumagen.

The errors of the inexpensive color probes you recommend can cause way more harm than good. Just ask manufacture technical support. I found the Eye-One's blue error at low levels to be quite lacking and the corrected picture overall to be a bit flat.

I'm waiting for Thx automated display calibration/standards to mature.
reincarnate is offline  
post #97 of 736 Old 05-13-2008, 03:00 PM
AVS Special Member
 
reincarnate's Avatar
 
Join Date: Jul 2004
Posts: 2,989
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
And now that I'm thinking, why not discuss how any UHP bulb changes its color temperature (greatly) as it ages. So a ISF (locked-in) calibration will need to be tweaked every few hundred hours.

I'm also waiting for LED light sources to eliminate the need for this constant recalibration. Why doesn't the ISF ever discuss this subject?
reincarnate is offline  
post #98 of 736 Old 05-13-2008, 10:49 PM
AVS Addicted Member
 
noah katz's Avatar
 
Join Date: Apr 1999
Location: Mountain View, CA USA
Posts: 20,346
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 81 Post(s)
Liked: 133
Tom,

"Ron, you ... appear to be operating exclusively on the level of subjective impressions, which OK as far as it goes."

"But since I can't see what you see, all I have to go on is the data. And the data has been remarkably consistent. Your subjective impressions simply aren't consistent with the data."

Tom, to put it bluntly, when will you get it through your head that WE DON'T CARE.

We want to like the image, not the numbers.

"You obviously really don't like it that I keep pointing this out, but why doing so is "unreasonable" is beyond me."

It's unreasonable because, again, you seem unable to grasp that your agenda of good objective numbers is not others' agenda of a good subjective image; if the numbers are bad and the image is good, it's fine with us.

So please stop telling us how bad Ron's, Lawguy's, and Dan's images much-improved images are.

Noah
noah katz is offline  
post #99 of 736 Old 05-14-2008, 03:44 AM
Advanced Member
 
Kevin McCarthy's Avatar
 
Join Date: Dec 2005
Posts: 729
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Noah: As long as everyone's weighing in here, I have to say that I'm with Tom on this one. I think he is fine with folks finding some combo of multiple knobs and levers that provides subjective improvement of the RS-1's color. What he has taken pains to point out (drawing a lot of flack in the process) is that the Vision gamut control, by itself, has distinct limitations and cannot properly correct the RS-1 color gamut errors. As a consumer on the fence myself, I really appreciate hearing this. I DO CARE. When Lumagen's President thanks him for his contributions and volunteers that "Tom is right that the Vision series does not provide gamut correction. Rather, it is gamut mitigation", what more is there to say?

Kevin
Kevin McCarthy is offline  
post #100 of 736 Old 05-14-2008, 04:35 AM
AVS Special Member
 
Lawguy's Avatar
 
Join Date: Sep 2004
Posts: 5,705
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Quote:
Originally Posted by Kevin McCarthy View Post

Noah: As long as everyone's weighing in here, I have to say that I'm with Tom on this one. I think he is fine with folks finding some combo of multiple knobs and levers that provides subjective improvement of the RS-1's color.

I don't know why I have to keep repeating myself on this point: it is not just subjective improvement. The end result measures objectively better. You can get results that measure better if you use the Vision, by itself, or the Vision in some combination with the RS1's color controls. Under most scenarios, the end result will be that your xy points are close to dead on, but the Y of the primaries will be slightly too low and the Y of the secondaries will be slightly too high.

The results are not perfect so it is true that it is not gamut "correction." But, I am convinced that had the RS1 been released with out of the box color like that we are getting with the Vision, no one would have complained about the RS1's color and, even though it is not "perfect" it is close enough so that most people would consider it to be an extremely accurate projector.

As for the Y of the colors being off, I can do a before and after comparison isolating the Y of each of the colors. If you want to do an experiment to roughly see how the Y of the colors is off, make two presents on your RS1, one on User1 and one on User2. On the first, have the color control on the JVC set to -5. On the second, have the color control set to 0. Throw up some color test patterns and look for the differences as you switch between the presets for each color. These are the kinds of lightness differences that we are talking about.

I agree that if you insist on measurably perfect gamut, buy a Radiance. Unfortunately, it costs nearly as much as a good new projector and I don't think that its cost justifies what it does. A Vision made sense for me because, for a modest investment, I was able to dial in perfect grayscale and gamma, and greatly improve the color gamut. Everyone has to do their own cost-benefit analysis on whether it is worth it.

I understand that Lumagen is planning a stripped down Radiance that has only one input and one output (HDMI only) that will be sold at a reduced cost. That seems like the product for me, but I suspect that it will not be released this year and probably not until much later next year.

Affable Nitwit
Lawguy is offline  
post #101 of 736 Old 05-14-2008, 05:23 AM
AVS Special Member
 
Ronomy's Avatar
 
Join Date: Aug 2004
Location: Massachusetts
Posts: 3,037
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 8 Post(s)
Liked: 18
Quote:
Originally Posted by Lawguy View Post

I don't know why I have to keep repeating myself on this point: it is not just subjective improvement. The end result measures objectively better. You can get results that measure better if you use the Vision, by itself, or the Vision in some combination with the RS1's color controls. Under most scenarios, the end result will be that your xy points are close to dead on, but the Y of the primaries will be slightly too low and the Y of the secondaries will be slightly too high.

The results are not perfect so it is true that it is not gamut "correction." But, I am convinced that had the RS1 been released with out of the box color like that we are getting with the Vision, no one would have complained about the RS1's color and, even though it is not "perfect" it is close enough so that most people would consider it to be an extremely accurate projector.

As for the Y of the colors being off, I can do a before and after comparison isolating the Y of each of the colors. If you want to do an experiment to roughly see how the Y of the colors is off, make two presents on your RS1, one on User1 and one on User2. On the first, have the color control on the JVC set to -5. On the second, have the color control set to 0. Throw up some color test patterns and look for the differences as you switch between the presets for each color. These are the kinds of lightness differences that we are talking about.

I agree that if you insist on measurably perfect gamut, buy a Radiance. Unfortunately, it costs nearly as much as a good new projector and I don't think that its cost justifies what it does. A Vision made sense for me because, for a modest investment, I was able to dial in perfect grayscale and gamma, and greatly improve the color gamut. Everyone has to do their own cost-benefit analysis on whether it is worth it.

I understand that Lumagen is planning a stripped down Radiance that has only one input and one output (HDMI only) that will be sold at a reduced cost. That seems like the product for me, but I suspect that it will not be released this year and probably not until much later next year.

The point is it still fixes the colors of all the primaries and secondaries so color itself is corrected. Only luma's are off. Frankly looking at a primary color and dropping it 30 % is hardly noticable on the screen. Very slightly less bright but at least the colors are as they should be.

By the way -5 color alone on the RS1 doesn't fix the colors where the Vision gamut does. I had to drop the JVC color down to -30 for the gamut to measure close to reference yet looking at real video it had very little color. The vision gamut fixes the color and luma is still kept high enough for good saturation.
Ronomy is online now  
post #102 of 736 Old 05-14-2008, 10:19 AM
AVS Addicted Member
 
noah katz's Avatar
 
Join Date: Apr 1999
Location: Mountain View, CA USA
Posts: 20,346
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 81 Post(s)
Liked: 133
""Tom is right that the Vision series does not provide gamut correction. Rather, it is gamut mitigation", what more is there to say?"

Exactly, but how many times does he need to say it?

Are there really any "consumers" who need constant protection in these threads?

Noah
noah katz is offline  
post #103 of 736 Old 05-14-2008, 01:40 PM
Advanced Member
 
Kevin McCarthy's Avatar
 
Join Date: Dec 2005
Posts: 729
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
I did.

Kevin
Kevin McCarthy is offline  
post #104 of 736 Old 05-14-2008, 10:11 PM
Senior Member
 
dhnjp1's Avatar
 
Join Date: Aug 2001
Location: escaped Silicon Valley
Posts: 408
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 13
I've only had a few minutes here and there the last few days, but I have finished some more experiments and understand the HDP/Q gamut adjustment limitations better now.

I tried 3 different approaches: the "balanced correction" approach, a "green only" approach, and a hit "exact" REC709 approach (R,G,B,C dE<3). All were done with RS1 color=0, and all Lumagen color controls at defaults.

All approaches run into one inherent limitation. The gamut corrections get smaller at lower IREs. In other words, all corrected primaries move closer to their original starting points as the IRE is dropped. The secondaries move by a different amount, which increases their error at lower IREs.

Even the balanced approach suffers from this problem. It is dropping the lumas by 30+%, not the "balanced correction" that minimizes, but does not eliminate, the problem. A "green only" correction gives the same secondary behavior as "balanced correction" if you drop all primaries' luma enough.

So in the end, we are back to subjective preferences. Objective "proof" is fine, but good grief, a .001 xy change can change dE by +3. No meter under $1000 is that accurate.

I prefer my "exact" REC709, as it is the most accurate at 100IRE, and the secondary shifts are not visible to me on test patterns, and certainly not on real video. The "balanced" method -30% luma is very obvious to me on 100% test patterns (looks like a 50% saturation pattern with gamut off), and is occasionally visible in real video as missing detail.

--Dan
dhnjp1 is offline  
post #105 of 736 Old 05-15-2008, 04:03 AM
AVS Special Member
 
Lawguy's Avatar
 
Join Date: Sep 2004
Posts: 5,705
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Up and until the latest firmware release, the Radiance suffered from the same problem at lower IREs. I try to calibrate at the 75% level. I find that it minimizes this effect and gives good results all around. What level are you taking measurements from?

When you have a moment, please post your preferred settings.

Thanks.

Affable Nitwit
Lawguy is offline  
post #106 of 736 Old 05-15-2008, 05:16 AM
AVS Special Member
 
Ronomy's Avatar
 
Join Date: Aug 2004
Location: Massachusetts
Posts: 3,037
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 8 Post(s)
Liked: 18
Quote:
Originally Posted by dhnjp1 View Post

I've only had a few minutes here and there the last few days, but I have finished some more experiments and understand the HDP/Q gamut adjustment limitations better now.

I tried 3 different approaches: the "balanced correction" approach, a "green only" approach, and a hit "exact" REC709 approach (R,G,B,C dE<3). All were done with RS1 color=0, and all Lumagen color controls at defaults.

All approaches run into one inherent limitation. The gamut corrections get smaller at lower IREs. In other words, all corrected primaries move closer to their original starting points as the IRE is dropped. The secondaries move by a different amount, which increases their error at lower IREs.

Even the balanced approach suffers from this problem. It is dropping the lumas by 30+%, not the "balanced correction" that minimizes, but does not eliminate, the problem. A "green only" correction gives the same secondary behavior as "balanced correction" if you drop all primaries' luma enough.

So in the end, we are back to subjective preferences. Objective "proof" is fine, but good grief, a .001 xy change can change dE by +3. No meter under $1000 is that accurate.

I prefer my "exact" REC709, as it is the most accurate at 100IRE, and the secondary shifts are not visible to me on test patterns, and certainly not on real video. The "balanced" method -30% luma is very obvious to me on 100% test patterns (looks like a 50% saturation pattern with gamut off), and is occasionally visible in real video as missing detail.

--Dan

Dan make sure COLR on the lumagen is set to 112. The factory reset sets it to 128 and that is wrong. You will see clipping in primary color ramps set to 128 and lose color detail. If you start with that you will get some wierd gamut settings and secondary shifting like you talk about.

I use the balanced method and do not see any shifting of gamut all the way down to 50IRE. At 30IRE the gamut returns to oversaturated but that's a know issue.

Remember with the gamut off and even with color set on the RS1 to -15 the gamut is still over saturated so comparing a color bar pattern to between gamut on and off isn't a good comparison. You would need an RS1 cal'd with a radiance to do a proper side by side evaluation.

In all my testing I couldn't see much of a difference in most material except that it eliminates the funny colors we sometime see.
Ronomy is online now  
post #107 of 736 Old 05-15-2008, 11:26 AM
AVS Special Member
 
TomHuffman's Avatar
 
Join Date: Jan 2003
Location: Springfield, MO
Posts: 6,357
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 15 Post(s)
Liked: 204
Quote:
Originally Posted by dhnjp1 View Post

So in the end, we are back to subjective preferences. Objective "proof" is fine, but good grief, a .001 xy change can change dE by +3. No meter under $1000 is that accurate.

Dan, this is a little hyperbolic. The worst case scenario in which one changes blue's y coordinate .001 increases dE by 2.1 Luv units. In all other cases, changes to either xy coordinate of just 0.001 results in a 1 dE or less change.

Also, I think the distinction between measured performance and subjective preference is a little overwrought--that is unless one just prefers inaccurate color, which evidently some people do. But for those people this entire enterprise is pretty much a waste of time in any case.

The dE method of color difference is not just an abstract theoretical model. It is also based on thousands of hours of empirical studies in which subjects stare at alternating color swatches and then report when they perceive changes in color.

There are some areas of performance in which measured results correlate poorly with perceived results. I just don't think that color is one of them. The standards for performance are precisely quantified and easily measured, though meter error does play a role here. I have found the Display 2 to be reasonably accurate, though as is often the case the more expensive i1Pro does better.

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

TomHuffman is online now  
post #108 of 736 Old 05-15-2008, 11:53 AM
AVS Special Member
 
Lawguy's Avatar
 
Join Date: Sep 2004
Posts: 5,705
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
This link suggest that the dE method is not a good way to evaluate color differences.

These are its conclusions:

Quote:


What are the two major findings?

Always when the color difference is due to a change in L* channel alone the perceptual difference is huge compared to any of the 10 cases where the color difference is due to a change in a* and/or b*. In other words the transfer function of the CIELAB is faulty. The dE(2000) formula increase this error the most.


None of the evaluated color difference models are in good agreement with our vision, or, more accurately, they are completely useless measures for color difference. E.g. dE=8 can mean that two colors are perceptually extremely different from each other but it can also mean that two colors are perceptually extremely similar.

It is interesting because it shows examples of colors at different dEs and you can see how much things can be different depending on a given dE.

If I understand this correctly, differences in luminance are especially noticable. As our Ys are all over the place, perhaps we need to think about this some more.

Affable Nitwit
Lawguy is offline  
post #109 of 736 Old 05-15-2008, 12:13 PM
AVS Special Member
 
Lawguy's Avatar
 
Join Date: Sep 2004
Posts: 5,705
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
After looking at the grapic dE charts that are found in the link above, I am kind of startled. A dE of 8 can look spot on to the reference while a dE of 1 can look off.

Can anyone explain what this means?

Affable Nitwit
Lawguy is offline  
post #110 of 736 Old 05-15-2008, 01:10 PM
AVS Special Member
 
reincarnate's Avatar
 
Join Date: Jul 2004
Posts: 2,989
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by Lawguy View Post

Tom,

If you, and others, are agreeable, I would like to start fresh. Why don't we do this in the correct way?

There are at least three of us that are trying this. The correct way to do this, is to start as you suggest. We should all post baseline measurements that represent out of the box performance. Why not also post settings with the color control at -15, but no other controls at work.

The next step is where the many paths diverge because there are several different strategies that appear to work. (1) Vision gamut controls only; (2) Vision gamut controls plus adjustments with other Vision controls (COLR, HUE, etc); (3) Vision gamut plus JVC color controls; and (4)All of the above (all vision and JVc controls).

Tom, if you are interested and the others agree, I would ask that we all work together under your direction. Tell us what to do and we will do it.

It really seems to me that this is all working. I don't know if we can get things to work any better than they are now, but it is possible.

What do you say?

A very nice and respectful post and probably the main purpose of AVS forum: for members to share information to get the best performance out of their gear.

Now why was it ignored? Why no sharing of data whatsoever? Because the ISF trained calibrators want to stay employed. (They will state that digital projectors vary too much, but this is only a quarter truth). Its that basic.
For example go read the Pioneer Kuro plasma owners thread. Calibrators shamelessly troll there for business.
Without being a customer, you usually have to stimulate them to get any type of response. See above. (in effect Huffman said umr was ignorant - There Will Be Blood?). They need to defend their reputation is it is important for their business.

Most members wisely kept their real world identity to themselves. Always question the neutrality of those who frequently post for their business. (manufactures excepted). As usual its all about the money.

If I want to be cynical (engineers frequently are) the push here is for people to spent $4K for the only 'good' video processor. One large technical problem is this $$$ processors dual HDMI outputs driving two displays, which for HDMI is ill-equipped and problematic to say the least. (HDMI was never designed for this important capability.) I'll bet that both HDMI outputs are identical and are limited to correcting only one of the displays at a time. The older DVI port solution is frowned upon by many when it should actually be preferred, as the audio can be stripped of by the receiver/AV controller first. Then a DVI based video processor (such as the HDP) can isolate and prohibit the second display's HDMI handshaking from interfering with the first displays response.

This was the only way to get the Monoprice 4*2 dual HDMI output switcher to work in my system. The first display has a Silicon Optix processor built in and the second display is driven by the Lumagen HDP. Now even my family can turn the power on an have both picture and sound appear without fiddling on both corrected displays simultaneously!
reincarnate is offline  
post #111 of 736 Old 05-15-2008, 02:33 PM
AVS Special Member
 
TomHuffman's Avatar
 
Join Date: Jan 2003
Location: Springfield, MO
Posts: 6,357
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 15 Post(s)
Liked: 204
Quote:
Originally Posted by Lawguy View Post

This link suggest that the dE method is not a good way to evaluate color differences.

It is interesting because it shows examples of colors at different dEs and you can see how much things can be different depending on a given dE.

If I understand this correctly, differences in luminance are especially noticable. As our Ys are all over the place, perhaps we need to think about this some more.

This page and its author (Timo Autiokari) is fairly well known, and I really don't know what to say, except that at least for this dataset, it sure seems that the L component plays a bigger role in color difference perception than hue and saturation differences. It also seems to show that the amount of color error for hue and saturation differences varies considerably depending upon the color involved and the background conditions.

Taking the dE 8 sample and dE94 method (which I prefer), for example the first swatch seems to show only very subtle differences when the hue and saturation are altered, but the next one shows more profound differences. But in both cases, it is where the L is altered (top and bottom right corners) where the differences seem most obvious.





As the page points out, CIE is constantly searching for more accurate ways to quantify color difference. The latest standard is CIECAM2002, which Microsoft has adopted for its Windows Color Management, but which the CIE has yet to officially endorse.

There is something else that occurred to me. All of the colors in this dataset are pastels. The 1976 system, on which the current dE system is largely based, substantially reduced the poor perceptual uniformity inherent in the 1931 system it replaced, but it did not entirely eliminate it. It may be that the residual lack of perceptual uniformity is especially obvious in pastel shades.

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

TomHuffman is online now  
post #112 of 736 Old 05-15-2008, 02:46 PM
AVS Special Member
 
Ronomy's Avatar
 
Join Date: Aug 2004
Location: Massachusetts
Posts: 3,037
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 8 Post(s)
Liked: 18
Quote:
Originally Posted by Lawguy View Post

This link suggest that the dE method is not a good way to evaluate color differences.

These are its conclusions:



It is interesting because it shows examples of colors at different dEs and you can see how much things can be different depending on a given dE.

If I understand this correctly, differences in luminance are especially noticable. As our Ys are all over the place, perhaps we need to think about this some more.

The only thing to think about then is to save up for a Radiance because it ain't happening with the vision series.
Ronomy is online now  
post #113 of 736 Old 05-15-2008, 02:49 PM
AVS Special Member
 
Lawguy's Avatar
 
Join Date: Sep 2004
Posts: 5,705
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
I think it means that you are better off having xy errors than Y errors. So, maybe get the best dE that you can with the smallest Delta Luma.

I would have thought the opposite would have been true and that xy errors would have been more noticable.

Affable Nitwit
Lawguy is offline  
post #114 of 736 Old 05-15-2008, 03:15 PM
AVS Special Member
 
Ronomy's Avatar
 
Join Date: Aug 2004
Location: Massachusetts
Posts: 3,037
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 8 Post(s)
Liked: 18
Quote:
Originally Posted by Lawguy View Post

I think it means that you are better off having xy errors than Y errors. So, maybe get the best dE that you can with the smallest Delta Luma.

I would have thought the opposite would have been true and that xy errors would have been more noticable.

The only way to get better Y numbers is to reduce the amount of gamut correction thus oversaturated colors return. I already tried this. Except for the oversaturated scenes returning all the other scenes looked the same as lower Y settings. DOH! I believe the errors are mostly in the higher IRE's and we don't see it because most scenes don't use a lot of color saturation anyway. Heck I can use settings that clip my primaries from 60 IRE and above and its hard to find a scene that has excessive loss of color detail and/or flat color. Video essentials you can see it because they wanted to use the entire color luma range.

Think about it...we only saw the bad colors in brighter scenes. Our eyes are just not as sensitive to color at lower IRE's and at lower IRE's the saturation is back anyway. We just tamed the upper IRE's.

I think if we want to use this gamut control you need to try and get the luma's balanced as close as possible. With my settings my red and green are only at 4 precentage points apart. Magenta is abot the same as green and the other two secondaries are higher but there isn't a thing I can do about it due to the excessive green correction.

Ron
Ronomy is online now  
post #115 of 736 Old 05-15-2008, 04:29 PM
AVS Special Member
 
Lawguy's Avatar
 
Join Date: Sep 2004
Posts: 5,705
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
I am not overly troubled by this at all.

I know that things measure quite well and I know that they look very good.

I wrote you earlier that I kept looking for real world references to try to determine how closely what I see on my display mirrors the real world. I think that those comparisons show that I am very close.

Now, if I place a can of Coke next to the screen, could I discern a difference in the color red from a can in a Coke commerical and the color of the can in the real world? I am not sure. But, things look as they should and I won't be bothered by the fact that perhaps I can see some subtle variation of the color as compared to a reference object that will never be present in my theater.

As much as there is a scientific-like procedure to go about a color calibration, there are so many weak links in the chain: questionable colorimeters, flaws in methods of measuring accuracy, flaws in the display and/or video processor. I find that I have to take a step back and look at what I see. It looks good to me. (don't get me started on the errors introduced in the film mastering process either).

I learned a lot about calibration in this process and I am glad that I did it because I significantly improved the quality of my picture.

I still question how much luma matters for us. I can move the color control on the JVC quite a bit before I see a noticable difference (and that would move the Luma of the colors beyond the range necessary to correct the Luma error for a given color).

So, I am happy.

Affable Nitwit
Lawguy is offline  
post #116 of 736 Old 05-15-2008, 04:42 PM
AVS Special Member
 
Ronomy's Avatar
 
Join Date: Aug 2004
Location: Massachusetts
Posts: 3,037
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 8 Post(s)
Liked: 18
Quote:
Originally Posted by Lawguy View Post


I still question how much luma matters for us. I can move the color control on the JVC quite a bit before I see a noticable difference (and that would move the Luma of the colors beyond the range necessary to correct the Luma error for a given color).

So, I am happy.

I feel the same! I am done with it and happy. Just enjoying the show now. I will be looking at the new lower port density radiance when I comes out. The HDQ has the port density I need. I don't need any more.

I enjoyed this but I put way too many hours on my lamp calibrating. Totally sick of calibrating and trying different settings. I am glad I created a 2.2 gamma along with my 2.4. I find that some shows on TV do look better with 2.2. They record some shows way to dark. I can see everything and it doesn't look washed out but why so dark? Then on another channel another show looks great at 2.4. Maybe the contrast levels they use are off I don't know.

Can't trust anything especially color on TV stations.

Ron
Ronomy is online now  
post #117 of 736 Old 05-15-2008, 08:37 PM
Senior Member
 
dhnjp1's Avatar
 
Join Date: Aug 2001
Location: escaped Silicon Valley
Posts: 408
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 13
Quote:
Originally Posted by TomHuffman View Post

Dan, this is a little hyperbolic. The worst case scenario in which one changes blue's y coordinate .001 increases dE by 2.1 Luv units. In all other cases, changes to either xy coordinate of just 0.001 results in a 1 dE or less change.

In one of my measurements, I changed one primary coord cell in the spreadsheet by .001 and the dE jumped from 12 to 15. That's when I realized that obsessing over only the dE numbers may not be very productive.
Quote:


Also, I think the distinction between measured performance and subjective preference is a little overwrought--that is unless one just prefers inaccurate color, which evidently some people do. But for those people this entire enterprise is pretty much a waste of time in any case.

I was actually agreeing with you Tom. The Vision series will never provide objective gamut improvement over anything beyond a narrow range of IRE. So we are left with subjective personal preference of the resulting image (saturation errors with perfect decoding vs. correct saturation with decoding errors).

In fact, I'm thinking that the whole concept of an external CMS box is doomed by the extreme complexity of external non-linear correction vs. the extremely simple linear correction that could be done in the display. One Teranex paper describes how doing non-linear gamut correction to 10-bit accuracy requires 100+ multiplications per pixel. Even the Radiance may never be capable of achieving the perfection you're looking for.

--Dan
dhnjp1 is offline  
post #118 of 736 Old 05-15-2008, 09:05 PM
Senior Member
 
dhnjp1's Avatar
 
Join Date: Aug 2001
Location: escaped Silicon Valley
Posts: 408
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 13
Quote:
Originally Posted by Ronomy View Post

Dan make sure COLR on the lumagen is set to 112. The factory reset sets it to 128 and that is wrong. You will see clipping in primary color ramps set to 128 and lose color detail. If you start with that you will get some wierd gamut settings and secondary shifting like you talk about.

I did have it at 112 for all of my recent measurements. But now I see that Jim Peterson posted in the Lumagen forum that 128 is the correct value for 1.0 gain, but some oversaturated displays like the RS1 look better with 112. I'll trust what Jim says since they can measure and compare the actual binary output with the input. That could explain why I've never been too bothered by the uncorrected RS1. The COLOR=112 would lower the brightness of all primaries.

Quote:


I use the balanced method and do not see any shifting of gamut all the way down to 50IRE. At 30IRE the gamut returns to oversaturated but that's a know issue.

With your original balanced settings, I can measure a small shift of the primaries and secondaries at 75IRE. Even though the magnitude is small, the direction of the shift is the same as for any settings without the big luma subtraction. Reducing all of the SUB* by an equal amount will increase the shift at lower IRE.

As an example of the magnitude of the shift with no luma reduction, my green-only correction gives y=.600 at 100IRE and y=.611 at 75IRE. Subtract some luma, and the magnitude of that shift decreases, but it is always there.

Quote:


Remember with the gamut off and even with color set on the RS1 to -15 the gamut is still over saturated so comparing a color bar pattern to between gamut on and off isn't a good comparison. You would need an RS1 cal'd with a radiance to do a proper side by side evaluation.

In all my testing I couldn't see much of a difference in most material except that it eliminates the funny colors we sometime see.

I'm more sensitive to the luma reduction. The first time I noticed it was a scene where a car pulls up to a sidewalk and brakes. With gamut off, I could clearly see the bright lightbulb inside the tail light. With balanced gamut on, I could barely see any increased brightness of the entire tail light, and could not see the actual bulb at all. I'm personally not willing to trade a loss of detail for color (xy) preference.

I haven't decided yet whether to keep gamut enabled or not. I'll watch a few more movies before doing anymore changes.

--Dan
dhnjp1 is offline  
post #119 of 736 Old 05-15-2008, 11:48 PM
AVS Special Member
 
TomHuffman's Avatar
 
Join Date: Jan 2003
Location: Springfield, MO
Posts: 6,357
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 15 Post(s)
Liked: 204
Quote:
Originally Posted by dhnjp1 View Post

In fact, I'm thinking that the whole concept of an external CMS box is doomed by the extreme complexity of external non-linear correction vs. the extremely simple linear correction that could be done in the display. One Teranex paper describes how doing non-linear gamut correction to 10-bit accuracy requires 100+ multiplications per pixel. Even the Radiance may never be capable of achieving the perfection you're looking for.

I am not really looking for perfection. That's an unrealistic goal. What I try to achieve is dE94 for RGBCYM at 1.0 or below measured at 75% stimulus. The Radiance gets you there.

But you are right in that color correction at the display level is much simpler. It's just that you can't count on the manufacturers to step up in this regard.

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

TomHuffman is online now  
post #120 of 736 Old 05-15-2008, 11:59 PM
AVS Club Gold
 
GlenC's Avatar
 
Join Date: Jun 2003
Location: Los Angeles - Whittier, CA
Posts: 2,555
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
When I am adjusting "Color", be it D65 grayscale or Primary/Secondary, I prefer to look at the "Color Difference" only:
((measured x - target x)^2 + (measured y - target y)^2)^.5, then, in addition, look at the measured x & y. Essentially, better to have a bigger error of +Blue than a small error +Green (within reason)...........

Delta E is really more appropriate for print than it is for video, IMO.

Glen Carter
Home Theater Calibration
www.ISFHT.com
GlenC is offline  
Reply Video Processors

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