JVC RS1/HD1 CC filter tweak - Page 5 - AVS Forum
Forum Jump: 
Reply
 
Thread Tools
post #121 of 132 Old 04-30-2007, 11:57 AM
AVS Addicted Member
 
krasmuzik's Avatar
 
Join Date: Oct 1999
Location: NewPort, VA
Posts: 11,270
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Hughman (Hugh2)

Usually RMS (root mean square) is used rather than average when you have negative values...statistically better. You only need to take RMS of dE since dE is the vector magnitude of dL, dC, dH already. I was going to update the spreadsheet - but Excel has no built in RMS - so it would be work to do later!

Have you tried it with the filter that bettered it for light colors at the expense of dark colors? The color control seems to compensate for the dark colors so it might be a matched tweak!
krasmuzik is offline  
Sponsored Links
Advertisement
 
post #122 of 132 Old 04-30-2007, 07:51 PM - Thread Starter
AVS Club Gold
 
Hughman's Avatar
 
Join Date: Jun 2004
Location: Canada
Posts: 2,247
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 12
Quote:
Originally Posted by krasmuzik View Post

Hughman (Hugh2)

Usually RMS (root mean square) is used rather than average when you have negative values...statistically better. You only need to take RMS of dE since dE is the vector magnitude of dL, dC, dH already. I was going to update the spreadsheet - but Excel has no built in RMS - so it would be work to do later!

Have you tried it with the filter that bettered it for light colors at the expense of dark colors? The color control seems to compensate for the dark colors so it might be a matched tweak!

Stats 101 is now starting to come back, it's been a welcomed very long time. I've updated my post to reflect a RMS calculation on dEu'v'.

I was hoping these results clarify things for me but it appears the opposite has occurred. It appears there can be a wide range of color control setting which achieve compromised but reasonably equal calibrated results. Think I'll need a few days to digest this.

My measured results were with the filter installed.


To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.
Hughman is offline  
post #123 of 132 Old 05-01-2007, 03:48 AM - Thread Starter
AVS Club Gold
 
Hughman's Avatar
 
Join Date: Jun 2004
Location: Canada
Posts: 2,247
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 12
I'll post the raw dE data for viewing for the filter calibrated -2 color setting which uses lightness as the determinant and the LCH model showing the best overall color compromise at a -10 color setting:

color control -2

........dEu'v'.....dL.....dC........dH
R......17.4.....2.3......17.0......3.2
G......28.2......-1.2......28.2......0.4
B......5.7......-1.0......4.5......-3.4
Y......20.9......0.4......20.5......-3.9
C......17.3......-0.6......16.9......3.7
M......7.1......1.6......6.8......-0.9

color control -10

........dEu'v'.....dL.....dC........dH
R......4.1......-1.7...... 2.1......3.1
G......25.0......-2.3......24.9......0.9
B......11.1......-4.7......-9.8......-2.6
Y......19.7......0.6...... 19.3......-4.2
C......15.9 ......-1.3......15.5...... 3.1
M......2.3......-2.1...... -0.9......0.6

I beg anyone with this PJ (JVC RS1) and measurement equipment to please perform similar measures and post your results. Not that this is life and death or anything but if other projectors behave in a similar manner then perhaps a range of color control settings could be found which could be effectively applied by any user asking how to best set the color control on the RS1 for maximum color fidelity. At this point I'm inclined to suggest a -10 color setting would be suitable at least for my system.


To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.
Hughman is offline  
post #124 of 132 Old 05-01-2007, 05:41 AM
AVS Club Gold
 
darryl b's Avatar
 
Join Date: Feb 2003
Location: wi
Posts: 709
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by Hughman View Post

... At this point I'm inclined to suggest a -10 color setting would be suitable at least for my system.

this is over my head.
do you mean to simply go to the image menu and set color to -10?
what do you mean by "suitable at least for my system?" is a filter required?
darryl b is offline  
post #125 of 132 Old 05-01-2007, 08:06 AM - Thread Starter
AVS Club Gold
 
Hughman's Avatar
 
Join Date: Jun 2004
Location: Canada
Posts: 2,247
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 12
Hi,

Yes, the color control I'm referring to is the one in the user "image" menu.

It has not been established yet but the 2nd wave pj's color control,which have the newer firmware, may operate differently than the first batch, therefore my control function may not be representative of all the RS1's out there. Without corroboration, my results are specific to my system only.

I did perform the measures with the filter installed and wil do a run without the filter soon but don't expect huge differences but ya never know.


To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.
Hughman is offline  
post #126 of 132 Old 05-01-2007, 12:13 PM
Senior Member
 
Stew M's Avatar
 
Join Date: Aug 2000
Location: Oregon Coast, USA
Posts: 327
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
I'm not as picky as many here, interested in some help for the obvious situations where oversaturation occurs. Not understanding all the posts here, is there some filter(s) that help make it better?

Quote:
Originally Posted by lovingdvd View Post

Unfortunately I don't think we have any cheap solution to the oversaturation issue.


Stew
The search for excellence is unending . . .
Stew M is offline  
post #127 of 132 Old 05-01-2007, 07:18 PM - Thread Starter
AVS Club Gold
 
Hughman's Avatar
 
Join Date: Jun 2004
Location: Canada
Posts: 2,247
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 12
I completed the color control measurements without the filter attached to the PJ. Though as I've stated above these measurements are relevant to my system only, however, with the many anecdotal reports of reduced color settings improving the perceived color balance I suspect other PJ's are reacting similarly. Both RMS and averge provide the same relative result so I'll only post the average but I've thrown in an additional measurement at -12.

-2, average dE=16.9
-4, average dE=16.3
-6, average dE=15.2
-8, average dE=14.7
-10, average dE=12.4
-12, average dE=14.6
-15, average dE=15.7


So regardless if the filter is installed or not the best color setting which minimizes perceived color error is at or around -10.


Stew M, the filters purpose has nothing to do with manipulating saturation levels. The filters purpose is to bring down excess blue and green more in line with red and do this optically instead of digitally with the PJ's RGB controls. This method of equalzing RGB levels increases contrast ratio. Sorry for the confusion, this thread is dealing with both topics.


To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.
Hughman is offline  
post #128 of 132 Old 05-02-2007, 06:20 AM
AVS Special Member
 
lovingdvd's Avatar
 
Join Date: Jul 2002
Posts: 8,887
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 52 Post(s)
Liked: 33
Hughman - great work on all this and thanks for the regular updates. The one question that keeps running through my mind as I read these posts is whether the error in Y, introduced by changing the color control, has a detrimental affect on PQ.

I haven't measured it recently but as I recall even changing the color control a few clicks lower caused Y to quickly become way off from the spec. So OTTOMH I'm thinking that at -10 the Y values are perhaps 20% too low - maybe more?

IIRC the compromise/tradeoff being made for this is based on finding a balance between 3 different properties (such as Y vs. CIE points vs. something else). My interpretation is that in this model all 3 factors are being equally weighted in importance. However, I do not know if that is indeed the case.

For example if Y has a more significant role than another factor, then I'm not sure a model that averages all 3 equally to find a compromise dE is ideal. Then again if it turns out Y is the least important factor, then that would be a big plus in favor of this tweak.
lovingdvd is offline  
post #129 of 132 Old 05-02-2007, 02:42 PM
AVS Addicted Member
 
krasmuzik's Avatar
 
Join Date: Oct 1999
Location: NewPort, VA
Posts: 11,270
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
dE is equal weighting of the perceptual components dL, dC, dH - in terms of it being a vector magnitude. So a 5% in lightness is just as annoying as a 5% hue or 5% chroma shift if that was the only error - the dE=5. However a 5% shift in all three is worse than a shift in each - the magnitude is now dE=8.6. (I sometimes use % rather than writing out delta units - it just reads better - though technically % is not a correct unit)

Keep in mind that we are using L* not Y. Y is the physical measure of brightness - that goes up quickly as a power curve - L* is the perceptual measure including the perceptual gamma - it goes up quickly in blacks then slows down in greys and whites. Basically it is like a mirror of the power curve - more like a square root curve. A huge change in absolute Y on the bright side is barely perceivable when you look at it as a perceptually relative L*. In fact a 20% Y cut is actually <9% cut in L*!


I think you will see those who adjusted per the CIE 1931 gamut chart and ignored Y are way out there at -20 and -30. But those who adjusted by eye - got the -10 that Hugh found thru perceptual measurement/analysis was the best compromise of gamut and decoding.
krasmuzik is offline  
post #130 of 132 Old 05-02-2007, 06:17 PM - Thread Starter
AVS Club Gold
 
Hughman's Avatar
 
Join Date: Jun 2004
Location: Canada
Posts: 2,247
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 12
Thanks Krasmuzik, I was hoping if I procrastinated long enough you'd offer up an understandable explanation.

Lovingdvd, the answers to your query have been posted in this thread and I'm hardly one to explain the mathematics so I'll refrain. If you require more info you'd be well served to do an LCH search under Krasmuziks name as he has a number of posts outlining the LCH model and it's advantages. Further up in this page you'll see a chart with dEu'v', dL, dC and dH numbers which should answer your brightness question.

You have the equipment, why not run a few measurements? I would welcome any measurements whether they corroborate my findings or not.

Krasmuzik, I'm getting the impression that using the LCH model to locate the best perceptible compromise between color saturation and brightness errors is a relatively new practice for display calibration. Was your inclusion of this tool influenced directly by your formal training or something you picked up on later on ?

What I need now is to have your spreadsheet or essence of integrated into my calibration software, I'm quickly running out of paper and pens.


To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.
Hughman is offline  
post #131 of 132 Old 05-03-2007, 12:14 PM
AVS Addicted Member
 
krasmuzik's Avatar
 
Join Date: Oct 1999
Location: NewPort, VA
Posts: 11,270
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Measurement of LCH to video calibration came from my own desire to create an objective perceptual coding system - since I and any other calibrator gets tremendous flack when we use subjective words to describe what we see - as usually those words will not agree with a PJ "champions" words when they have self or biz justification blinders on. This became evident after I conducted a series of shootouts with local AVSers - and made the mistake of asking them to post their impressions - some of whom have never posted again as a result.

LCH is not covered as well as CIELUV though - so it took a number of references for me to piece together the math. Bruce Lindblooms and Poynton reference actually has LCh - where h is degrees or radian angle - it is not linked to the dE formula as it just the polar coordinate form of L*u*v* (what is the magnitude of an angle?!).

I found what I needed in an obscure Color Science Matlab reference - that gave the math for H so that LCH and LUV result in the same dE. Sometime after that I saw gregr started using u'v' gamut charts at WSR - and I sent him an early version of the math for review. He noted that he is not convinced u'v' is any more perceptually correct than xy so he flip-flops which chart he uses. He was right - the more perceptual version is actually u*v* - which is only a target - not a gamut - especially in video with RGB components that are standard specified for unequal brightness.

I have a Digital Pictures text from my computer engineering masters degree that covers a lot of psychovisual science as background for the compression algorithms. It covers the various forms of CIE charts - xy, u'v' and u*v* and how the Munsell representation of color maps to them - and shows how u*v* is the most perceptually uniform. Though it is still not perfect since Munsell is a color circle notation and there is no measurement system that results in perfect circles - for surface colors CIELAB has been tweaked since 1976 (with tweaks that diverge based on industry) but CIELUV (used for colored light/video) has not been tweaked. But I had no intention of inventing anything new - this spreadsheet is stock CIE 1976 formulas.

So I started testing it on my calibration customers since gregr himself seemed ambivalent about it and thru my own education knew it still ain't perfect!. I came up with my own chart formats thru that process with the goal being I don't want to explain color science and video engineering so they can understand what calibration did for them. It is thru that process that I began to appreciate how closely it maps to human perception. Once I finalized the presentation thru feedback on calibration jobs I started posting calibration reviews on my website using CIELUV and CIELCH.

ISF does not teach any of this color science - nor do advanced calibration seminars. Even today you don't see xyY measures for color in reviews - let alone the more perceptual LCH version - I would be happy if we get them to stop meaningless 6500K charts (the only failing ISF question - why is 6500K not a calibration measure when D65 is?) Red push never even seems to be measured - it seems to be adjusted thru AVIA filters or reference viewing! The most annoying is reviewers that show the same CIE gamut charts for the same PJs - then have completely different subjective interpretations of how off they are! That is where my subjective grading method came from - but I tried to make it as tolerant as possible (my personal version would be much more harsh!) - using dE RMS for final grades and assigning ranges for subjective qualifiers - but you still need the LCH breakdown to explain the errors.

I posted the spreadsheet to educate regarding the math - I just don't want to see people start selling it by plugging in colorimeter readers. If they want to use the spreadsheet to create such a product - they cannot. What they can do is go back to the references and learn it themselves and do their own implementations, but even then my chart format is my copyright - they need to come up with their own designs. I reserve the right to make money from my work and not allow anyone else to - they have to do their own work if they want to make money.

ColorFacts has LCH in its raw data window - but never really worked for me. It also has u'v' gamut diagrams - it is missing the L*u*v* and LCH color targets for decoders and gamuts that indeed would be useful in calibration - though it really has to be a wizard since these sort of adjustments requires retaking the RGBCMYW readings each time - as you have learned! I basically just use ColorFacts for its link to sensors/generators and dump out raw data for my own software. If someone has a reference CMS and wants their color calibrated - I charge extra for using my methods instead of the usual AVIA filters and reference viewing - they are getting more than an ISF certified calibration after all!

BTW I don't use this spreadsheet in my own calibrations. I reserve the right for there to be bugs in it as a result - I already found one and posted an update. There could be more - I only slapped it together as a result of the recent interest in color measures because of the JVC hype - it is not something I use myself - it is only an educational reference. Like all good textbooks - it may be revised.
krasmuzik is offline  
post #132 of 132 Old 05-04-2007, 06:18 AM - Thread Starter
AVS Club Gold
 
Hughman's Avatar
 
Join Date: Jun 2004
Location: Canada
Posts: 2,247
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 12
Wow, very interesting. Thanks for the reply. When I mentioned the spreadsheet being used in a program I had in mind that you would be paid for it's use if used commercially.


The spreadsheet is most definitely educational and imo a very useful tool. While ideally there would never be a need to make compromises, PJ's such as the JVC make it clear that indeed some compromise may be necessary and luckily with this pj can be made to to maximize the perceivable result. The LCH model has potential to be a valuable tool when used for this purpose and I'm surprised and a little dismayed that some would dismiss the method. Calibrating the chain is the game, if compromise is required I see nothing incorrect with using a psychovisual approach to finding the best compromise as it is a human brain at the end of the chain. To ignore or dismiss outright a seemingly objective tool which can ultimately be used for this purpose seems a little contrary to the ultimate goal of a calibrator. IMO, your method is a far more reasonable and enlightened approach than simply chanting the mantra and claiming it can't be done.


To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.
Hughman is offline  
Reply Digital Hi-End Projectors - $3,000+ USD MSRP

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