3D LUT Calibration - Page 3

X-Rite was no help. I asked if dark readings are required every ten minutes for TV calibration and here is their response:

X-RITE doesn't support TV calibration unless it is being used as a monitor. X-RITE only issues a White Calibration tile with the i1Pro.

That's all they had to say.

Given the near three hour calibration example above you would not interrupt to do a meter calibration? Is this 2 hour and 47 minute total time the definition of "calibration run"?

I know my questions sound very obvious to you, Steve, but calibrating with LUTs is very foreign to me and X-Rite documentation regarding TV display calibration is non-existent.

Is LUT pronounced with a long "a" as in salute?

For display calibration, are the images stored in the software?

Are the images what is referred to as "calibration flats" or are the images "patches"? What is the difference?

Is computer HDMI output to display HDMI input correct or is there something more involved? EDIT - referencing displaying images to be measured.

Would a workflow look something like the following?

1. Set up the display including black and white levels
2. Do a quick profile
3. Measure (as in http://lightillusion.com/hubblemanual.htm)
4. Create LUT(s)
5. Transfer to box
6. Hook up = Source > Box > Display


Note: HDMI will be required for the box despite licencing fees.

From the ChromaPure calibration software manual:

There is no commercially available CMS solution for consumer gear that can correct non-linear performance throughout the entire color space."

Is this a definition of the gap a LUT box will fill?

Oops, I accidenty said "long A". My mistake. Slut, gut, mutt, LUT. Got it.

I've been through your web site three times and it's coming together, very slowy.

If the calibration images are from a source other than the LightSpace software, how does the suppying device communicate with LightSpace for the "time per frame" input and how does LightSpace know which image is being displayed?

Is it correct, then, that the only items required are a PC, LightSpace, Meter, and the future LUT box?

Costs indeed. In post #21 I came up with $2488 USD for a fully functional licence and I realize that LightIllusion cannot easily break out that which would be required for TV display calibration only and eliminate all the other abilities of the LightSpace software so that it could be sold at a lower price.

Sell me three or four 3D LUT boxes and charge me a reasonable fee for a two day license so I would have time to build LUTS for three or four home televisions and I'm probably a LightIllusion customer.

Furthermore, for professionals, $2488 USD could be recovered in short order by a calibrator or installation company. A single sale might yield a $100 retail markup on a LUT box and another $300 or so for LUT building. Then, when the customer buys a replacement TV, $200 - $300 more for the required new LUT building. In addition, with a fast meter like the Hubble, a customer call would take two hours instead of five, effectively doubling the number of calibrated displays in any given time period and thus doubling the total profit.

There would be opposition by the customer to the costs indicated in the above paragraph versus the modern day CalMan/ChromaPure $300 calibration utilizing the controls on the display. Reuseability of the LUT box would help soften the blow.

Being the first in the marketplace would enable you to make the LightIllusion label a household word. Television manufacturers use THX and ISFccc identification as much for marketing as they do for picture improvements. Get LightIllusion circuitry installed inside television sets and you will rule the calibration world. Get in line to be the 2013 Time Magazine "Man Of The Year".

One question though, and it's a big one - next post

This LightIllusion web page is very convincing.

The apples and oranges picture: The 1D vcgt RGB gamma LUT picture looks dumbed down. Did you hunt down a worst case cross color error display to make your example or can the kind of 3D vs 1D LUT picture improvement shown on the web page be expected on an average television display in my home?

I love the capability, but cost is a huge factor for us home consumers.

My Epson 1080UB projector has a full CMS, custom gamma..., but it isn't implemented properly, so I have to make compromises. Would love the ability of this 3D LUT box. However, it would need to have enough bit depth to prevent banding or other artifacts and also not delay the signal noticeably.

Also, how would your 3d LUT box differ from the TruVue eeColor Processor which has 3D LUT capability and bundled software to run it.

Dan

Perhaps it is my computer monitor. The 1D vcgt RGB gamma LUT picture looks relatively dull and the 3d LUT picture looks stunning. I try to put it into perspective: Displays that I have calibrated look very good and I truly cannot visualize them able to get that much better. I obviously need to see it for myself.

Thanks for the reply. Just so I am clear with your reply, are you saying that the Truvue processor is capable of adjusting primary and secondary Hue Saturation and Lightness to correct the gamut, but doesn't take into account any cross color or linearity problems?

Dan

It appears that the initial image is the 3D LUT and mousing over produces the 1D LUT, the opposite of the text on the page - "The initial image is a 1D LUT and when you hover your mouse over it the image will show the result of a 3D Cube, both LUTs being based on exactly the same data."

Nope the second image is the 3dlut.

The 3dlut is also not correcting for the D55 to D65 twist for 10bit log.Marcie is still too blue. Also in the 1dlut case marcie appears to have been converted to video using the kodak log to lin conversion numbers with a 2.2 gamma added ...these numbers are not for video conversion but for normalising the data for the straight line portion of the curve for linearisation with a simple 1.7 gamma , so you get black and white clipped below 95 and above 685.

And the massive difference you see in saturation is because its targeting a print gamut from the original 10bit log. This example is somewhat irrelevant to home cinema calibration. No one involved in this thread is going to be grading 10bit log film scans into video.

If you are only targeting a rec.709 display target the value of 3dluts over the other CMS methods is extremely small and you would likely be operating in tolerances that are beyond the vast majority of sensors available at this price range.

Other than the quote you're talking way over my head, Mr.D.

If a 3D LUT has little value beyond the results obtainable in a display containing 10 pt. grayscale/gamma and full CMS, would it not have much more value calibrating a display with little more than 2 pt. grayscale or a difficult to calibrate display such as the Samsung HL67A750 LED DLP which requires many trips in and out of the service mode to calibrate correctly?

I was and still am hoping to see some large difference as indicated on this LightIllusion page containing the following statement:

All consumer orientated calibration systems, such as i1-Match, iColor, CalMan, HCFR, and the prosumer orientated programs such as EqualEyes, are all vcgt based, and are therefore very limited in their ability to manage colours because vcgt's are 1D RGB LUTs, limiting their colour control capability, as described elsewhere within this website (1D LUTs can't control saturation (gamut) or cross colour elements).

But, as these systems do not tend to explain this limitation they are assumed to be accurate display calibration techniques, which they are most definitely not. The following image is the same frame shown with a 1D RGB (vcgt) display profile, compared to a real 3D LUT. As can be seen the difference is large. The calibration used is obviously not valid for YOUR display, but the differences between the two LUTs is valid.

The first image initially shows the original frame, and when you mouse over will show the effect of a 1D vcgt RGB gamma LUT - the best any of the above described calibration systems can provide.

Yes I am aware of that , assuming anything else would be a bit absurd. I have a nice 4k film scan of marcie data on the computer I'm typing this on (actually its 3656x1556 because marcie originated as academy and was scanned on a Kodak Cineon scanner about 1993...did I mention I'm very familiar with the image?)


Yes but I can still tell what the baseline targets are by looking at the result. Its the 10bit log "Marcie" film scan with a 3dlut of a film print stock model baked into it with an additional video gamma of some description as the baseline. (its obviously not log or linear baseline) What it hasn't done is convert the D.55 balance that marcie and most 10bit log film scans have to the D.65 display balance of a notional video display paradigm.

Yes its a 3dlut conversion because of the massive and typical impact the print gamut had on saturation. Yes its some sort of video tonescale correction because in terms of contrast its kinda okay for a video display although the 1dlut has been converted badly using the loglin numbers as the basis of the video conversion as I've said , however I can tell its not corrected the colour balance from print to video with the 3dlut version because Marcie is too cold looking.

This could mean the baseline display target was already balanced to D.55 and didn't require the 3dlut to correct for it ( I do this myself occasionally). I have also done it for linear display baseline with and without the D.55 to D.65 twist. However its a bit dumb then comparing it with a 1dlut version that does have the D.55 correction.

The reason I can tell this is that I know very well indeed what Marcie is supposed to look like as I've been looking at it for 15 years in a variety of colorspaces and a variety of display luts. Did I mention I worked for Kodak...in the precise department that generated the Marcie LAD chart...in fact?


First of all its "Marcie" not "Mary" and her real name isn't actually "marcie" anyway (did I say I was very familiar with the image and where it originated)

Secondly as I design and maintain color pipelines as part of my work using a wide variety of CMS systems from many different companies over the years including 3dlut based systems before anyone else was using them (its kinda easy to get accurate print stock models when its actually Kodak that is doing it), Heard of the Kodak Display Manager for example?

Only the other month I was designing LUTS for viewing 10bit log film scans on video displays ( sRGB baseline in this case) using ...you guessed it...marcie as an additional visual verification. So I can absolutely tell when marcie has been converted to video with a 1d or 3dlut and I can also tell if the D.55 balance has been corrected at the same time.

Ironically in that example image the 1dlut has been corrected from D.55 to D.65 and the 3dlut hasn't been corrected hence its too cold as I've said before.

Now this is all with reference to displaying a 10bit log film scan on a video display so the obvious difference in gamut representation is completely meaningless for a home user discussing the pros and cons of 3dlut correction vs other forms of CMS with a view to calibrating a video display ( call it rec.709 for arguments sake).

Do i think 3dluts are useful and accurate ...absolutely for relevant applications.

Do I think a 3dlut profile based system is going to be noticably superior to a CMS that affords multi point gamut correction for targeting a rec.709 environment on consumer gear...No I don't.

Do I think automated profile based calibration is a more sensible approach ...been saying that for years and years.

I also don't think the home cinema arena needs an overpriced lut box shackled to an exorbitant software license for something as simple as a 3dlut builder. If you can do it for under $500 without some carpetbagger software licence scheme I wish you the best of luck...bearing in mind people have done the same thing for PCs ...for free and/or with much more functionality than mere 3dlut CMS.


No need to shout...and as I've explained its not a particularly valid comparison in its own terms and its of little relevance to calibrating home cinema displays. ( or any video display for that matter).

Just like to add that "Equaleyes" is only the simple graphics hardware LUT loader of the cinespace suite. Its never been represented as anything other than that . The cinespace suite itself contains routes to 3dlut building and application. Equaleyes and cinespace are definitely not "prosumer".

Tell me, Mr.D, from your excerpted quotes above it looks very much like Steve's project might be just the ticket for my two examples below, eh?

It sounds like you are hankering after a more automated calibration than necesserally the 3dlut loveliness for its own sake.

3dluts are great. However I really get bored of people treating them like the second coming..they are what they are no more or less.

No, Mr.D, you're the one negating the "loveliness" in the first quote. I merely asked if the proposed LightIllusion consumer 3D LUT system will give a 10 point grayscale/gamma and full 3D CMS quality calibration to displays that do not possess these controls. I already knew the answer and was waiting for your confirmation but instead I received, ummm, "attitude"....

Oh attitude is it ? Apologies I'll write it in crayon for you next time.

Any 3dlut profile based system will give as good a correction as possible within its design parameters. That does not automatically mean its better worse or the same as another CMS implementation that does not require full 3d colorspace calculations and transformations.

An offboard CMS will also give you controls that your display hardware does not possess ...again its a question of how they are designed and implemented that matters.

(Facile hint : the fact that something uses 3dluts does not automatically make it superior)

However please yourself and insist on needlessly defending a bit of kit that does not actually exist and in my opinion will probably not come to fruition from Light Illusion anyway. Cue lots of passionate denial and outrageous indignation for suggesting such a thing...before the thread goes to sleep and the product doesn't appear anyway ....

Incidentally Steve how many individuals actually make up Light Illusion?