Is it possible to calibrate a display using a high quality digital camera? - AVS Forum
Forum Jump: 
Reply
 
Thread Tools
post #1 of 208 Old 09-16-2010, 02:27 AM - Thread Starter
Advanced Member
 
morphinapg's Avatar
 
Join Date: Jun 2006
Posts: 605
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 12
I realize it wouldn't be nearly as accurate, but would it be possible to calibrate the greyscale of a display using a high quality digital camera that is calibrated to D65 greyscale? I would place the camera's lens right up next to a grey image, take a picture using the calibrated camera, and then determine if the RGB levels all match (or are close). If not, I would tweak my display's settings until they do. Is this possible? I know it wouldn't be as accurate as other methods or be the proper way to do things, but shouldn't a camera white-balanced to D65 show as grey on a calibrated display, or at least near it?

NOTE: This thread was never intending to suggest using the digital camera as a replacement for a proper calibration. This was simply an experiment I wanted to try, to see if my display's color could be improved using a digital camera.

EDIT: Procedure for those who wish to try the experiment:

before starting, note your TV's original settings

STEP 1
Set your camera's white balance as close as you can to D65. I used a sheet of paper in overcast daylight (wikipedia said overcast is better than direct daylight).

STEP 2
Display an image filling the screen with a shade of grey. Which shade(s) you choose is up to your calibration strategy.

I personally used a 50% grey to find the best color temperature preset, and then a 25% screen for RGB cutoff and a 75% screen for RGB drive, compromising between the two settings.

STEP 3
Find a good exposure for the shade of grey you are shooting. Make sure that no RGB levels in the image reach 255 (I use photoshop's levels option to make sure there are no levels reaching 255), to ensure that the image is not overexposed, which can distort your color. You will likely end up with a negative EV as EV 0 is usually too bright for these type of measurements.

STEP 4
Take your pictures. Get the camera's lens as close to the screen as you can, and defocus the lens as much as you can. Using lower resolutions is fine, as we're only looking for one color. Open the pictures in your favorite image editor. Crop the images by 50% (in case the edges of the image are darker than the rest) and then average the remaining pixels (in photoshop, filters/blur/average). If you don't have an average filter, other blur filters might work, or you can take multiple samples and average them yourself.

STEP 5
Sample the image's color. After the image has been averaged, take a sample of one of the pixels to determine the image's color. Grey has RGB levels that are equal. It doesn't matter what number the RGB levels are (since different exposures change that), it just matters that they are equal.

STEP 6
Adjust your TV's settings accordingly. If there is too much red, either decrease red, or increase blue and green. If there is too much green, either decrease green or increase red and blue. If there is too much blue, either decrease blue or increase red and green. Perform the opposite if there is too little red, green, or blue.

STEP 7
Repeat steps 4-6 until that shade of grey appears as close to pure grey in your image editor as you can get. Note your changed settings.

STEP 8
Repeat steps 4-7 for as many grey levels as you want, to ensure greyscale looks good from dark to light. I personally tweaked both my dark and light levels separately, and then compromised between the two settings.

STEP 9
Repeat steps 4-8 until you have achieved the best greyscale you can achieve for each level. This may take several attempts for each level, as changing your settings for one level is likely to change the accuracy for another.

STEP 10
Re-calibrate your brightness and contrast, as these are likely to have been changed by your modifying the RGB controls of your display.

STEP 11
Note your new settings, and switch back and forth between your new settings and your old settings. If your new settings appear worse than your old settings, reset your settings to their original values. If you are still unsure, watch a show or two with your new settings. If you notice anything strange with your new settings (some things are too green, etc) then switch back to your old settings.

Note: All cameras are different. This may or may not help you. If you can, use other methods to improve your display. This is in no way intended as a substitution for a proper calibration.
morphinapg is offline  
Sponsored Links
Advertisement
 
post #2 of 208 Old 09-16-2010, 09:31 AM
AVS Special Member
 
sotti's Avatar
 
Join Date: Aug 2004
Location: Seattle, WA
Posts: 6,605
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 21 Post(s)
Liked: 167
no it doesn't work.

The camera simply isn't designed to work that way. Many have tried, all have failed.

Joel Barsotti
SpectraCal
CalMAN Lead Developer
sotti is offline  
post #3 of 208 Old 09-16-2010, 12:04 PM - Thread Starter
Advanced Member
 
morphinapg's Avatar
 
Join Date: Jun 2006
Posts: 605
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 12
Quote:
Originally Posted by sotti View Post

no it doesn't work.

The camera simply isn't designed to work that way. Many have tried, all have failed.

I understand that it may not be designed to work that way, but surely with a color balance set to D65, all other processing disabled, and a manual exposure set by the user, it could be done, right?
morphinapg is offline  
post #4 of 208 Old 09-16-2010, 12:44 PM
AVS Special Member
 
airscapes's Avatar
 
Join Date: Dec 2008
Location: Philadelphia
Posts: 4,612
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 50 Post(s)
Liked: 109
Try it and then have the display calibrated with the proper equipment and see how close it was.
airscapes is offline  
post #5 of 208 Old 09-16-2010, 12:49 PM
AVS Special Member
 
sotti's Avatar
 
Join Date: Aug 2004
Location: Seattle, WA
Posts: 6,605
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 21 Post(s)
Liked: 167
Quote:
Originally Posted by morphinapg View Post

I understand that it may not be designed to work that way, but surely with a color balance set to D65, all other processing disabled, and a manual exposure set by the user, it could be done, right?

No because the camera itself isn't accuratly calibrated.
Nor is it designed to filter the light according to the standard XYZ observer function.

Joel Barsotti
SpectraCal
CalMAN Lead Developer
sotti is offline  
post #6 of 208 Old 09-16-2010, 01:58 PM
AVS Club Gold
 
SierraMikeBravo's Avatar
 
Join Date: Mar 2007
Location: Topeka, KS
Posts: 1,429
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 19 Post(s)
Liked: 31
OP, do a search. I think you'll find that this subject has come up more often than it should. Best wishes!

Shawn Byrne
Erskine Group
HAA Design Certified -THX Certified Professional

Design-Video & Audio Calibration Information

The original Pro Theater Layout
SierraMikeBravo is offline  
post #7 of 208 Old 09-16-2010, 02:40 PM - Thread Starter
Advanced Member
 
morphinapg's Avatar
 
Join Date: Jun 2006
Posts: 605
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 12
Quote:
Originally Posted by sotti View Post

No because the camera itself isn't accuratly calibrated.
Nor is it designed to filter the light according to the standard XYZ observer function.

The camera is calibrated to a daylight color balance. It may not be 100% perfect, but it's close enough for me. I don't really know anything about XYZ in terms of color, just RGB.

I plan to eventually get the right tools to calibrate, but for now I thought this would be a fun experiment, and possibly allow me to achieve similar results. I'm not saying this should be as good as a real calibration, but should hopefully bring me closer than with no other tools.
morphinapg is offline  
post #8 of 208 Old 09-16-2010, 03:59 PM
AVS Special Member
 
sotti's Avatar
 
Join Date: Aug 2004
Location: Seattle, WA
Posts: 6,605
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 21 Post(s)
Liked: 167
Quote:
Originally Posted by morphinapg View Post

The camera is calibrated to a daylight color balance. It may not be 100% perfect, but it's close enough for me. I don't really know anything about XYZ in terms of color, just RGB.

I plan to eventually get the right tools to calibrate, but for now I thought this would be a fun experiment, and possibly allow me to achieve similar results. I'm not saying this should be as good as a real calibration, but should hopefully bring me closer than with no other tools.

You're just as likely to get good results putting your display in warm 2.

It'll give you practice calibrating, but it's not even really an educated guess.

Joel Barsotti
SpectraCal
CalMAN Lead Developer
sotti is offline  
post #9 of 208 Old 09-16-2010, 07:48 PM - Thread Starter
Advanced Member
 
morphinapg's Avatar
 
Join Date: Jun 2006
Posts: 605
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 12
Quote:
Originally Posted by sotti View Post

You're just as likely to get good results putting your display in warm 2.

It'll give you practice calibrating, but it's not even really an educated guess.

I did end up starting with Warm 2 (as that provided the closest greyscale without tweaking), but I ended up tweaking the various RGB settings in the service menu until I got a greyscale that only resulted in 2-3% saturation in both light and dark areas. After that, I watched a couple movies on the set and I found that after this "calibration" I much preferred the image to how it was before I "calibrated" it.

I'm using "calibration" in quotes because I don't want anybody to think I'm recommending this as a replacement for professional calibration. That's not what I'm doing at all. However, I do think it can produce better results than simply using your eyes to judge it. Although, it is important that you have a high quality digital camera that allows you to set everything manually, especially color balance, because otherwise you will be calibrating your set to the wrong color for grey.
morphinapg is offline  
post #10 of 208 Old 09-16-2010, 08:41 PM
Member
 
herzzreh's Avatar
 
Join Date: Sep 2005
Posts: 83
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
If you have a high-quality digital camera, it means your computer monitors are calibrated, which means you're more than capable of calibrating your TV. Same principle applies.
herzzreh is offline  
post #11 of 208 Old 09-17-2010, 09:42 AM
 
ChrisWiggles's Avatar
 
Join Date: Nov 2002
Location: Seattle
Posts: 20,730
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 14
As many previous threads have dealt with this the answer is pretty straightforward:

No, you can't. Sorry.
ChrisWiggles is offline  
post #12 of 208 Old 09-17-2010, 11:27 AM - Thread Starter
Advanced Member
 
morphinapg's Avatar
 
Join Date: Jun 2006
Posts: 605
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 12
Quote:
Originally Posted by ChrisWiggles View Post

As many previous threads have dealt with this the answer is pretty straightforward:

No, you can't. Sorry.

I did, and like I said, I'm not looking for an entirely accurate method. While that would be nice, it's unrealistic to expect a digital camera, which hasn't been designed for calibration to provide perfect results. However, I expect it to provide better results than the default values on the set. After calibrating with the camera, I did find that the picture looked better with my new settings.

I eventually plan to get the tools necessary to do a proper calibration, but until then, this is the best I can do.
morphinapg is offline  
post #13 of 208 Old 09-17-2010, 12:09 PM
 
ChrisWiggles's Avatar
 
Join Date: Nov 2002
Location: Seattle
Posts: 20,730
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 14
You are better off just guessing by eye.

You cannot use a digital camera as anything more than an optical comparator.

I don't mean to be harsh about this, but when you say that "I'm not looking for an entirely accurate method" this implies that it remains a method that is at least marginally accurate in some way. Let me be blunt: it is not. I don't mean to berate this point, but a camera is NOT designed to be colorimetric tool. Even a basic tristimulus meter inherently requires 4 different sensors and filter curves that match the CIE standard observer. A digital camera only has 3 types of photosites (RGB) and it is impossible to get any kind of accurate objective measurement of color.

As I mentioned, you can use a camera to compare colors on one display, but in order to target an actual desired color, you need another measurement device that is capable of doing that. Which begs the question: if you have the capability of measuring color int he first place, why bother with the camera.

If you want to use the camera merely as an optical comparator to achieve a linear grayscale, then that is a limited but appropriate use of the tool. But in terms of actually targeting a color for white, you might as well just pick a white by eye because what the camera sensor "sees" and yields as equal portions RGB is an entirely arbitrary result that has no connection to any kind of actual CIE color. It's really no different than putting each of the R, G, and B gain controls at the same numeric value on your TV so that they are all "equal." Those numbers have nothing to do with anything in terms of achieving an accurate D65 neutral white. Using a camera is simply the illusion of using a tool or measuring something when it is not measuring anything at all, and telling you nothing at all useful about the color actually being produced.

It would be akin to measuring the size of an object in inches, and then judging how much it weighs based entirely on its size and volume. It's simply impossible for that to work, and is not providing you any information at all about how much the object actually weighs. It's not just that measuring how big something is is somehow a less accurate way to do it. It's not a way to do it in the first place, because it CAN'T tell you anything without more information. Same goes with the camera method. You can't use it as a tool unless you have a WHOLE LOT more information. It is simply not a tool at all.

I'm sorry to disappoint you, and I understand the desire to figure out an enterprising and DIY way to do something (and I encourage it) but you also have to understand what it is you're actually doing and whether it will even work. This is very much unequivocal: it does not work, will not work, and suggesting that it sort of maybe is halfway effective is totally incorrect.
ChrisWiggles is offline  
post #14 of 208 Old 09-17-2010, 12:18 PM - Thread Starter
Advanced Member
 
morphinapg's Avatar
 
Join Date: Jun 2006
Posts: 605
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 12
Quote:
Originally Posted by ChrisWiggles View Post

You are better off just guessing by eye.

You cannot use a digital camera as anything more than an optical comparator.

I don't mean to be harsh about this, but when you say that "I'm not looking for an entirely accurate method" this implies that it remains a method that is at least marginally accurate in some way. Let me be blunt: it is not. I don't mean to berate this point, but a camera is NOT designed to be colorimetric tool. Even a basic tristimulus meter inherently requires 4 different sensors and filter curves that match the CIE standard observer. A digital camera only has 3 types of photosites (RGB) and it is impossible to get any kind of accurate objective measurement of color.

As I mentioned, you can use a camera to compare colors on one display, but in order to target an actual desired color, you need another measurement device that is capable of doing that. Which begs the question: if you have the capability of measuring color int he first place, why bother with the camera.

If you want to use the camera merely as an optical comparator to achieve a linear grayscale, then that is a limited but appropriate use of the tool. But in terms of actually targeting a color for white, you might as well just pick a white by eye because what the camera sensor "sees" and yields as equal portions RGB is an entirely arbitrary result that has no connection to any kind of actual CIE color. It's really no different than putting each of the R, G, and B gain controls at the same numeric value on your TV so that they are all "equal." Those numbers have nothing to do with anything in terms of achieving an accurate D65 neutral white.

I first calibrated the camera's white balance to daylight (using a custom set white balance, not a preset). It may not be a perfect representation of D65, but it's pretty close. I then used the calibrated camera to take pictures and level out the RGB levels.

I'm not trying to take the same kind of measurements that the meters do, I'm just trying to use the camera as a workaround for now.
morphinapg is offline  
post #15 of 208 Old 09-17-2010, 12:53 PM
 
ChrisWiggles's Avatar
 
Join Date: Nov 2002
Location: Seattle
Posts: 20,730
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 14
Quote:
Originally Posted by morphinapg View Post

I first calibrated the camera's white balance to daylight (using a custom set white balance, not a preset). It may not be a perfect representation of D65, but it's pretty close. I then used the calibrated camera to take pictures and level out the RGB levels.

You're not listening.

Even if you have a perfect D65 reference to align the camera to, the SPD of a display that matched D65 may appear entirely different to the camera sensors.

You can't compare one SPD at D65 to another SPD at D65 with a camera because you have no idea whether it will "measure" the same to the camera at all, and in all likelihood will "measure" quite differently.

Quote:


I'm not trying to take the same kind of measurements that the meters do, I'm just trying to use the camera as a workaround for now.

You're missing the point. It's not that you're taking different, or less accurate, or inferior measurements.

You aren't taking a measurement of ANYTHING at all.

I don't know how else to reiterate that. You aren't measuring anything.

You are picking entirely arbitrary numbers that have nothing to do with anything. As I said the first time, you might as well guess by eye. At least you have some assurance that it appears sort of white to you visually.
ChrisWiggles is offline  
post #16 of 208 Old 09-17-2010, 01:01 PM - Thread Starter
Advanced Member
 
morphinapg's Avatar
 
Join Date: Jun 2006
Posts: 605
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 12
Quote:
Originally Posted by ChrisWiggles View Post

You're not listening.

Even if you have a perfect D65 reference to align the camera to, the SPD of a display that matched D65 may appear entirely different to the camera sensors.

You can't compare one SPD at D65 to another SPD at D65 with a camera because you have no idea whether it will "measure" the same to the camera at all, and in all likelihood will "measure" quite differently.

I don't know what SPD means, but why wouldn't two colors that are exactly the same show as exactly the same on a camera? Thats what cameras were designed to do, to accurately represent the colors of our world. I'm not talking some cheap camera from wal-mart, I'm talking high quality digital cameras which have manual control of everything.

Quote:
Originally Posted by ChrisWiggles View Post

You're missing the point. It's not that you're taking different, or less accurate, or inferior measurements.

You aren't taking a measurement of ANYTHING at all.

I don't know how else to reiterate that. You aren't measuring anything.

You are picking entirely arbitrary numbers that have nothing to do with anything. As I said the first time, you might as well guess by eye. At least you have some assurance that it appears sort of white to you visually.

They are measurement of what the camera sees. The camera sees in RGB, just like the human eye. Why wouldn't it be representing accurate color when it has been calibrated to do so? And why wouldn't the meters used in calibration suffer the same type of problems?

Plus, you can't do this by eye, because your eyes can adapt to a wide range of colors and see them as true white, that doesn't mean it is.
morphinapg is offline  
post #17 of 208 Old 09-17-2010, 04:23 PM
AVS Special Member
 
Kilgore's Avatar
 
Join Date: Nov 2004
Location: Huntsville Ontario
Posts: 2,988
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 60 Post(s)
Liked: 171
Let me get something straight here.

Are you trying to convince the myriad of experienced and extremely knowledgeable calibrators here that you CAN use a digital camera to calibrate a display, or are you asking the myriad of experienced and extremely knowledgeable calibrators here to help you understand why you can't?
Kilgore is offline  
post #18 of 208 Old 09-17-2010, 04:35 PM - Thread Starter
Advanced Member
 
morphinapg's Avatar
 
Join Date: Jun 2006
Posts: 605
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 12
Quote:
Originally Posted by Kilgore View Post

Let me get something straight here.

Are you trying to convince the myriad of experienced and extremely knowledgeable calibrators here that you CAN use a digital camera to calibrate a display, or are you asking the myriad of experienced and extremely knowledgeable calibrators here to help you understand why you can't?

Using the word "calibration" loosely, I'm saying it is possible, and that I have done it. It may not be close to a real calibration, and it's probably outside of the range of error real calibrators deal with, but I believe it has gotten me a result more accurate than I could have done simply with my eyes, and most definitely better than the stock settings on the set.

A camera is a device that is designed to mimic human eyes, so why wouldn't it be possible? A camera sees in RGB, a human sees in RGB. As long as the camera is white-balanced to the correct white point (I used daylight, which is pretty close to D65), it should theoretically be possible to calibrate a tv. The camera should be a high quality camera where everything can be set manually of course.

If I am wrong, I am asking the experts here to explain exactly why, given all the above information it is not possible? Assuming I only know what I have stated that I know, explain to me why it wouldn't work.
morphinapg is offline  
post #19 of 208 Old 09-17-2010, 05:08 PM
AVS Special Member
 
Kilgore's Avatar
 
Join Date: Nov 2004
Location: Huntsville Ontario
Posts: 2,988
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 60 Post(s)
Liked: 171
Quote:
Originally Posted by morphinapg View Post

If I am wrong, I am asking the experts here to explain exactly why, given all the above information it is not possible? Assuming I only know what I have stated that I know, explain to me why it wouldn't work.

"If I am wrong"

If...?

EDIT: ChrisWiggles knows what he's talking about. I would listen to what he's trying to explain to you. You simply are lacking the technical knowledge to understand what he's saying (no offense).
Kilgore is offline  
post #20 of 208 Old 09-17-2010, 05:35 PM
 
ChrisWiggles's Avatar
 
Join Date: Nov 2002
Location: Seattle
Posts: 20,730
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 14
Quote:
Originally Posted by morphinapg View Post

I don't know what SPD means, but why wouldn't two colors that are exactly the same show as exactly the same on a camera?

First, I just want to start out that I really don't mean to be argumentative about this, and I hope you don't interpret my posts that way, despite their vehemence.

I got my start in this entirely out of my own passion as a DIYer and I am always supportive of people who are trying to learn to do something and figure something out on their own, so please don't interpret this as someone who is "defensive" about some kind of purported professionalism or something (unfortunately a common thing in the A/V world).

SPD= spectral power distribution.

I'm going to move very quickly through this explanation because I have dinner cooking on the stove and it's friday night, I have other things to do.

The CIE color chart describes the way that humans see color. Humans have three receptors for color, but their curves overlap and are complex:

http://en.wikipedia.org/wiki/File:CI..._Functions.svg



Note how the x color matching function for the human standard observer has two portions of response across the visible spectrum.

These curves interact with the particular spectral power distribution of any light source that falls within the visible spectrum. The way these three receptors respond and interact with the exact distribution of light yields basically a sort of "total" amount for R, G, and B curves that then map to a particular color that we percieve.

Pay attention closely to that last part: the particular curve or particular spectral power distribution (an infinite number of values that describe how much light is present at each particular wavelength through the visible spectrum) is sensed only in broad swaths by our eyes/brain. So for any particular CIE color that we perceive, there are an infinite number of physical stimuli that can all have totally different spectral power distributions. So consider another viewer, say an animal that has only 1 receptor, or has more than 3 receptors than we have, or has 3 receptors that have entirely different response curves to what we think of as Red Green and Blue, they may be able to distinguish colors differently than we do. They may either not see differences between SPDs that we can, or they may see differences in SPDs that we cannot.

A digital camera only has 3 receptors, whose curves do not match the sort of 4-portion curves of our receptors. The particular response curves of any given camera can vary a lot, which is why cameras have different kinds of color response. None are fundamentally designed as colorimetrically accurate: they don't match or describe CIE color as perceived by humans.

The way a camera reacts to different SPDs can be unpredictable. You can for instance have two displays right next to each other, calibrated with a proper instrument to an accurate D65 reference, which will inherently appear 100% identical to the human eye. However, the actual SPD of those displays may vary considerably (because there are an infinite number of SPDs that yield a particular CIE color result, hence there are an infinite number of possible SPDs that are all the same color as D65 despite being entirely different physical stimuli, and whatever the light source or particular panels of that particular display will vary from display to display. CRT phosphors have very different than say an LED-backlit LCD panel, etc). The camera may "see" those two SPDs as different, perhaps substantially different, because the camera won't have the same response curves as the human visual system because the camera wasn't designed to. And it will "see" those two different SPDs as different, despite the fact that to the human eye they are entirely identical colors, with exactly identical CIE color coordinates. This is the fundamental reason for color metamerism, or color metamers, which are different SPDs which appear identical, and this is fundamental to the human visual system and the range of unique colors we are both capable of distinguishing as distinct, and incapable of distinguishing as distinct.

On the other hand, a different situation may also arise where two displays with very different colors displayed on them, which appear visually different to the human eye, may be "seen" as totally identical to a given camera, again depending entirely on the interaction between the particular SPD of the display and its interaction with the particular response curves of the camera.

This is why cameras are not in any way accurate or useful to describe actual CIE color. They are not designed to, nor are they capable of doing so. You can only really use a camera in very limited circumstances to do color work like this if you understand exactly what its filter curves are, AND you know what the SPD is of the light source you're measuring is which involves profiling the camera very carefully. This requires very expensive spectroradiometry equipment which is going to cost you many many thousands of dollars more than the camera. You have to know what you're doing, and it has to be very controlled. And again you would be using the camera in concert with equipment that is designed and capable of determining CIE color exactly and in an objective manner.

I would love to say that you could just point a camera at something and measure color based on its results. But you simply cannot.

And it is important for people to understand that in a very fundamental way. I would hate to see this idea gaining a foothold and spreading like a virus, as have many other very misunderstood"good-ideas" in the A/V world that rest entirely on not really understanding what it is one is doing.


Quote:


Thats what cameras were designed to do, to accurately represent the colors of our world. I'm not talking some cheap camera from wal-mart, I'm talking high quality digital cameras which have manual control of everything.

No, they are not. This is something you don't have the background to understand, and I hope the explanation above begins to shed some light on exactly WHY a camera doesn't see the same way the human eye sees. You could design a camera to see that way in theory, but I am not aware of any cameras that do so, because it would be a very proprietary imaging system with at minimum 4 primary sensors, rather than 3.


Quote:


They are measurement of what the camera sees. The camera sees in RGB, just like the human eye. Why wouldn't it be representing accurate color when it has been calibrated to do so? And why wouldn't the meters used in calibration suffer the same type of problems?

Again, because the human eye sees RGB in a very particular manner which differs from a camera.

Calibration meters come in two basic varieties, and some do suffer similar problems. "Tristimulus meters" do something what you expect them to do, they basically give you a gauge of how much R, G, and B there is with a stimulus in a manner that matches the human eye. But as I described above, because of the particular XYZ curves in the CIE system, this actually requires usually 4 minimum sensors with filters that are designed to come as close as possible to the CIE curves. They differ in their ability to properly closely match those curves, and with particular SPDs their precision can be questionable.

The second variety operates nothing at all like a camera: a spectroradiometer. This is usually a series of hundreds of photosites behind a diffraction grating which breaks out the light into a rainbow of colors spatially (think shooting light through a prism). This measures across the entire visible spectrum in very small chunks of wavelengths and gives you the exact SPD of the physical stimulus, from which is calculated the CIE color based on the human viewer curves. These meters can cost tens of thousands of dollars, and are measuring the exact physical stimulus beyond simply the human response to it (a tristimulus colorimeter is only attempting to "see" the colors that humans see).

Quote:


Plus, you can't do this by eye, because your eyes can adapt to a wide range of colors and see them as true white, that doesn't mean it is.

Exactly. Doing it by eye is pretty arbitrary. But so is doing it with a camera. You are using an entirely inappropriate device that does NOT match the CIE standard observer curves, and will interact with the particular SPD of what you're measuring in an entirely unknown and predictable way.

With a proper colorimeter you have strong confidence, or near-absolute confidence (with a spectro-meter) of what the actual CIE xy coordinates of a particular color are. You can say with certainty what a particular color actually IS.

When you point a camera at something, you are getting entirely arbitrary values of arbitrary "R" "G" and "B" primaries that do not relate in any way to defined scientific values and cannot be used to derive anything about CIE color. You can't say anything about the color itself with any certainty. All you can say is that this particular camera reacted to that particular stimulus in some arbitrary way that in NO WAY relates to any defined values within color science. A different camera may react entirely differently. And the same camera may react entirely differently to the same color from a different display with an entirely different SPD. That's why you CANNOT use a camera for this task in the way you want to.

I wish you could, but you can't. Please understand this basic fact. I'm not just naysaying a good idea. It's a good idea only to people who haven't studied color and don't understand the way the human visual system reacts to light in the way that we see color.

If you come up with an easy way for a camera to see in a CIE-accurate way easily, then you stand to make a FORTUNE.

And I hope some others will chime in too, because I know there are several people around here who have more experience with color science than I do and can probably write a more scientifically accurate version of what I attempted to summarize here.
ChrisWiggles is offline  
post #21 of 208 Old 09-17-2010, 06:26 PM
 
ChrisWiggles's Avatar
 
Join Date: Nov 2002
Location: Seattle
Posts: 20,730
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 14
Quote:
Originally Posted by morphinapg View Post

Using the word "calibration" loosely, I'm saying it is possible, and that I have done it. It may not be close to a real calibration, and it's probably outside of the range of error real calibrators deal with, but I believe it has gotten me a result more accurate than I could have done simply with my eyes, and most definitely better than the stock settings on the set.

You have not performed a calibration to an objective reference, and it is not possible to measure color with a camera in the way you desire.

What you have done is laudable, and the creativity and thought, given lack of background in color science is great. BUT, that doesn't mean what you did was correct or useful!

The entire purpose of using a measurement device is to MEASURE something using some defined value. We measure length with rulers using inches or meters. These are defined scientific values that actually mean something. We measure speed using meters per second. Again, SI values that mean something. We can measure energy, volume, etc again all using defined scientifically accepted values.

"Red" is not a value. "Green" is not a value. "Blue" is not a value. These terms have absolutely no scientific meaning. They relate to no physical stimulus and have no objective meaning whatsoever.

When you point a measurement device and attempt to align a display to D65 which is neutral white, D65 is a defined coordinate (and actually a defined SPD as well) that objectively characterizes a particular physical stimulus that is unambiguous and can be entirely characterized using physics. In other words, our end-goal is a scientific VALUE with actual meaning, that is well defined. And the tool we use to measure whether it sees such an exact simulus must be CAPABLE of telling us whether or not a particular color of gray on a monitors is or is not D65. And a camera is NOT capable of doing that.

It would be like measuring distance with an elastic band with markings on it rather than a measuring tape. How many "inches" you measure with an elastic band depends entirely on how tightly you pull the band. Maybe you happen to pull it so that you get exactly the right measurements by accident. But the likelihood of you doing that is slim.

Quote:


A camera is a device that is designed to mimic human eyes, so why wouldn't it be possible?

Because a camera is NOT designed to mimic the human eyes. This is a fundamental point that you clearly were not aware of.

Now you are aware of this fact. Hence it follows why it is not possible.


Quote:


A camera sees in RGB, a human sees in RGB. As long as the camera is white-balanced to the correct white point (I used daylight, which is pretty close to D65), it should theoretically be possible to calibrate a tv. The camera should be a high quality camera where everything can be set manually of course.

Again, nope.

Quote:


If I am wrong, I am asking the experts here to explain exactly why, given all the above information it is not possible? Assuming I only know what I have stated that I know, explain to me why it wouldn't work.

The above post is why you are wrong. When people describe the human visual system as seeing in RGB, that's not exactly wrong, but it's a HUGE oversimplification. "Red" isn't a thing that we see. There is no such thing as "RED" light, really. What you think of as "red" is really a curve that describes the response of a photoreceptor in your eye to a particular range of frequencies of light. The same goes for "green" and "blue." And unless the curves on a camera are EXACTLY the same as the curves in your eyes, then the "red" that a camera sees may be different. That's the fundamental issue. And because the curves overlap, and one of them has a double-dip (actually a negative portion which is modeled with that second part because otherwise you run into the problem of negative-light), you need minimum 4 sensor curves (four primaries in the camera) to capture that. Which is why "tristimulus meters" don't have the three sensor types implied by the name, but instead usually a minimum of 4. Some have more than that. And spectrometers have hundreds, but they are trying to measure much more than simply colors distinguished by the human eye...
ChrisWiggles is offline  
post #22 of 208 Old 09-17-2010, 09:32 PM - Thread Starter
Advanced Member
 
morphinapg's Avatar
 
Join Date: Jun 2006
Posts: 605
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 12
Thank you ChrisWiggles for that large amount of information, lol. I don't think I understand it all right away, but I understand the overall idea. One question though, you say the human eye has 3 sensors, yet these meters need a minimum of 4. Why is that? Can't the meters have 3 sensors just like the eye? If so, why do cameras use a different curve (not really sure what the curves mean other than they are different in the human eye vs the camera) instead of simply using a curve that closely matches the human eye?

By the way, I know the RGB "measurements" I'm making are not referring to a specific CIE color, but I was under the assumption that if the camera was calibrated to D65, that any "true grey" image would have RGB levels that were equal. I get now that two different displays with the same CIE color, but with different SPDs (still not entirely sure what that really means) will show up as a different color. I guess my question is, how different will those colors potentially be. For example, if a color is mostly green on one display on the camera, could that same color potentially be mostly blue on a different display on the camera?

I wouldn't think there would be a huge difference like that, but if there is, then this idea is most definitely a bad idea. So far, my TV seems better off because of it, so maybe my camera is more accurate than you expected for my TV's SPD, or maybe I just had good luck, or maybe my eyes are just fooling me lol.

In any case, this has certainly given me experience in a calibration, even if the result of that calibration isn't what a real calibration would provide.
morphinapg is offline  
post #23 of 208 Old 09-18-2010, 03:06 AM
 
ChrisWiggles's Avatar
 
Join Date: Nov 2002
Location: Seattle
Posts: 20,730
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 14
Quote:
Originally Posted by morphinapg View Post

Thank you ChrisWiggles for that large amount of information, lol. I don't think I understand it all right away, but I understand the overall idea. One question though, you say the human eye has 3 sensors, yet these meters need a minimum of 4. Why is that? Can't the meters have 3 sensors just like the eye? If so, why do cameras use a different curve (not really sure what the curves mean other than they are different in the human eye vs the camera) instead of simply using a curve that closely matches the human eye?

As in the image above, notice how the x curve has two parts. It's not a single smooth curved area, there is a whole other section that overlaps entirely with the blue area. Making an optical filter like that is not really feasible, so they use two different filters in front of two photosites to get that whole curve in two parts.

Good cameras and good films generally try to get a decent approximation of the human eye, but the metric there is more about what looks good, what looks "right" and artistic intent. It's not about scientific measurement of actual color perception. It's just about making a good-looking image. And there's all KINDS of slop that happens, for instance nearly any digital camera can see IR no problem, but our eyes cannot. And this can affect some shots in significant ways. And it can certainly affect how a camera reponds to light if we want the camera to be an objective measure.

Quote:


By the way, I know the RGB "measurements" I'm making are not referring to a specific CIE color, but I was under the assumption that if the camera was calibrated to D65, that any "true grey" image would have RGB levels that were equal.

RGB levels of what?

This is what the SPD of D65 actually looks like:
http://upload.wikimedia.org/wikipedi...d0/SPD_D65.png

Quote:


I get now that two different displays with the same CIE color, but with different SPDs (still not entirely sure what that really means) will show up as a different color. I guess my question is, how different will those colors potentially be. For example, if a color is mostly green on one display on the camera, could that same color potentially be mostly blue on a different display on the camera?

Not a whole lot different. But when we're talking about measuring color with precision, we're talking about knowing with good certainty what a color's coordinates actually are. Not just that it's sort of "green" looking as opposed to "red" or something totally different.

Quote:


I wouldn't think there would be a huge difference like that, but if there is, then this idea is most definitely a bad idea. So far, my TV seems better off because of it, so maybe my camera is more accurate than you expected for my TV's SPD, or maybe I just had good luck, or maybe my eyes are just fooling me lol.

There won't be huge differences. But the way any camera sees "greens" or "reds" or whatever will vary depending entirely on the camera. That's why they all look different. It's why film stocks all can look different. Not usually in major ways, but in subtle ways. And in ways that far exceed what any proper measuring device for achieving accurate color should yield.

And when you deploy something like a camera that can only tell you that something is "green-ish" or "really green" or "not-so-green" with any certainty, that doesn't really help you in calibrating. You're just as well off looking at the green and asking yourself: does this look way too greenish? Or not enough? Does this look pretty neutral for white? Or not so much? The camera is likely to get you in the ballpark for white. But then so are your eyes. But then you've only gained entrance to the ballpark, and there are a LOT of seats in a ballpark, and they're all in different places.

Quote:


In any case, this has certainly given me experience in a calibration, even if the result of that calibration isn't what a real calibration would provide.

It's all a learning process, I hope you find this more an encouragement to learn more, rather than a defeat. Color is fascinating, and I'm only just scratching the surface myself. What I don't know yet about color is astounding!
ChrisWiggles is offline  
post #24 of 208 Old 09-18-2010, 03:54 AM - Thread Starter
Advanced Member
 
morphinapg's Avatar
 
Join Date: Jun 2006
Posts: 605
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 12
Quote:
Originally Posted by ChrisWiggles View Post

As in the image above, notice how the x curve has two parts. It's not a single smooth curved area, there is a whole other section that overlaps entirely with the blue area. Making an optical filter like that is not really feasible, so they use two different filters in front of two photosites to get that whole curve in two parts.

Okay, I guess that makes sense, but I still don't really understand why the curve has two parts, or what the curve even really represents.

Quote:
Originally Posted by ChrisWiggles View Post

Good cameras and good films generally try to get a decent approximation of the human eye, but the metric there is more about what looks good, what looks "right" and artistic intent. It's not about scientific measurement of actual color perception. It's just about making a good-looking image. And there's all KINDS of slop that happens, for instance nearly any digital camera can see IR no problem, but our eyes cannot. And this can affect some shots in significant ways. And it can certainly affect how a camera reponds to light if we want the camera to be an objective measure.

That's true, and maybe the IR problem is one of the reasons different calibrated displays would show up so differently on the same calibrated camera, maybe they have different IR levels.

Quote:
Originally Posted by ChrisWiggles View Post

RGB levels of what?

This is what the SPD of D65 actually looks like:
http://upload.wikimedia.org/wikipedi...d0/SPD_D65.png

Not really sure how to interpret that SPD, but when referring to RGB levels, assuming the camera is properly calibrated, here is what I know: Every real color can be represented at a large range of values, but if the exposure is too high, values will be clipped and that will reduce accuracy. So for the most accurate results, we need to be sure that all values of RGB are far below 255. Now, for example pure red has an RGB value of 255/0/0, but can be represented anywhere from 1/0/0 to 255/0/0, so the best way to represent that color is to take the highest value in the RGB values, and assign that as 100%, and then scale the other values accordingly. In this example, regardless of the R value (for example, 58/0/0), the result is 100%/0%/0%.

For another example, magenta might be captured on camera as 40/0/40, but when converted to percentage, that's still 100%/0%/100%, which corresponds to pure magenta. If my RGB reading reads 69/12/116, then I scale that to 59%/10%/100%, which corresponds to a purplish color. Now, if you understood everything above, you would realize that a pure grey should have level RGB values, which would correspond to 100%/100%/100%. I tweak the RGB settings in the service menu until my readings get as close to pure grey as possible. If any readings show RGB values at 255 (or even close to that, to be safe) I adjust the ISO, F-Stop, shutter speed, or EV value (depending on different circumstances) until the RGB values are low enough for my satisfaction.

Quote:
Originally Posted by ChrisWiggles View Post

Not a whole lot different. But when we're talking about measuring color with precision, we're talking about knowing with good certainty what a color's coordinates actually are. Not just that it's sort of "green" looking as opposed to "red" or something totally different.

And that's fine, but as I said in my initial post, I wasn't looking for precision at this time, just a step closer than stock settings.

Quote:
Originally Posted by ChrisWiggles View Post

There won't be huge differences. But the way any camera sees "greens" or "reds" or whatever will vary depending entirely on the camera. That's why they all look different. It's why film stocks all can look different. Not usually in major ways, but in subtle ways. And in ways that far exceed what any proper measuring device for achieving accurate color should yield.

And when you deploy something like a camera that can only tell you that something is "green-ish" or "really green" or "not-so-green" with any certainty, that doesn't really help you in calibrating. You're just as well off looking at the green and asking yourself: does this look way too greenish? Or not enough? Does this look pretty neutral for white? Or not so much? The camera is likely to get you in the ballpark for white. But then so are your eyes. But then you've only gained entrance to the ballpark, and there are a LOT of seats in a ballpark, and they're all in different places.

I think I can trust my camera more than my eyes, as I know for a fact I've gotten used to sets that were way too warm or way too cool, without even noticing it until they were fixed for me. Most people, unless the colors are obviously off, won't see their sets looking bad initially, but will appreciate their sets much more once they have been calibrated.

Quote:
Originally Posted by ChrisWiggles View Post

It's all a learning process, I hope you find this more an encouragement to learn more, rather than a defeat. Color is fascinating, and I'm only just scratching the surface myself. What I don't know yet about color is astounding!

I agree, and I am certain I will continue to research this and I plan to eventually know how to calibrate my sets the right way lol but for now, at least this is good practice, and hopefully my set has slightly better color, even if it's just a small amount better than it was. Eventually I'll make it even better.
morphinapg is offline  
post #25 of 208 Old 09-18-2010, 05:29 AM
AVS Special Member
 
airscapes's Avatar
 
Join Date: Dec 2008
Location: Philadelphia
Posts: 4,612
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 50 Post(s)
Liked: 109
Wow, .. um.. hum.. Normally when multiple experts tell me I can not do something, I accept it. But on some occasions I just can not wrap my head around what they are telling me, so I do scientific tests to SEE what they are talking about.

Morphinapg, if you really want to understand and see why they say this does not work, I would suggest hiring a pro to come to your house with his $10,000 meter and software package and have him calibrate your display. He will educate you as he is doing the work, assuming you pick a good calibrator. He will take post calibration measurements of how you now have your set configured with your camera, then reset to defaults for the setting you started from and take more measurements, then do an actual calibration.
Once complete you will have a properly calibrated display, and reports with graphs to show you how the gray scale and color was with your camera settings, factory setting and the proper settings. Why the camera didn't work will no longer matter, the fact that your tv can really look that good and why don't the manufactures set it up this way to start with will be more of what you will be thinking about.

Seeing is believing and no matter what is discussed here you will not believe it until you see it!
Good luck!

PS I do not do calibrations but have read enough to know that doing it ones self is a hobby and I would rather know it is correct and watch, then play with my display. I just spent $400 to have my 5 year old 720p projector calibrated (projector is maybe worth $400) and the gains in picture and color quality was well worth the money! I am not rich, I do not have a dedicated theater and it was painful to part with that cash! It was also well worth it and I will enjoy the image for the next 800-1000 hours of lamp life.. yeah I know it will shift as the lamp ages but man is it nice now!
airscapes is offline  
post #26 of 208 Old 09-18-2010, 12:40 PM
 
ChrisWiggles's Avatar
 
Join Date: Nov 2002
Location: Seattle
Posts: 20,730
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 14
Quote:
Originally Posted by morphinapg View Post

Okay, I guess that makes sense, but I still don't really understand why the curve has two parts, or what the curve even really represents.

Because the response of the receptors (CIE colod matching functions) in our eyes looks like this:

http://en.wikipedia.org/wiki/File:CIE1931_RGBCMF.svg


Quote:


That's true, and maybe the IR problem is one of the reasons different calibrated displays would show up so differently on the same calibrated camera, maybe they have different IR levels.

Only one reason, but it is among them yes.

Quote:


Not really sure how to interpret that SPD, but when referring to RGB levels, assuming the camera is properly calibrated, here is what I know: Every real color can be represented at a large range of values, but if the exposure is too high, values will be clipped and that will reduce accuracy. So for the most accurate results, we need to be sure that all values of RGB are far below 255. Now, for example pure red has an RGB value of 255/0/0, but can be represented anywhere from 1/0/0 to 255/0/0, so the best way to represent that color is to take the highest value in the RGB values, and assign that as 100%, and then scale the other values accordingly. In this example, regardless of the R value (for example, 58/0/0), the result is 100%/0%/0%.

Those are arbitrary choices of signal levels that are defined by video standards, not color science. Then you have signals that are equal in amounts of R' G' and B', then video standards define that as a neutral white, and most imaging standards (but not all) define neutral white as D65. But neutral white could just as easily be something else, like D55, or 9300K, or whatever the standard dictates.

In color science, CIE Illuminante E is the equal-energy illuminant, which has uniform power across the visible light spectrum. That's more the physical "neutral white" that you might be looking for, but that is not the neutral white point chosen by video standards and that is implied by values that are equal in R'G'B' values in a video source. D65 doesn't have equal energy of light across the spectrum, and looks significantly different than the E point, as you saw in the SPD chart of D65 I posted above.

Maybe the camera sensors will give you equal amounts of R'G'B' for the E point. Or for some other arbitrary color temperature. And it will depend entirely on the SPD of whatever the camera is pointed at. So just because a camera is giving you equal amounts of R'G'B' values tell you nothing really about the color that the camera is sensing.


Quote:


For another example, magenta might be captured on camera as 40/0/40, but when converted to percentage, that's still 100%/0%/100%, which corresponds to pure magenta. If my RGB reading reads 69/12/116, then I scale that to 59%/10%/100%, which corresponds to a purplish color. Now, if you understood everything above, you would realize that a pure grey should have level RGB values, which would correspond to 100%/100%/100%. I tweak the RGB settings in the service menu until my readings get as close to pure grey as possible. If any readings show RGB values at 255 (or even close to that, to be safe) I adjust the ISO, F-Stop, shutter speed, or EV value (depending on different circumstances) until the RGB values are low enough for my satisfaction.

But you're not actually tweaking anything. Do you understand that? Let me ask you a question instead: when you are tweaking your R'G'B' values so they are the same off the camera, what is your target? When you get them all equal, what white have you now arrived at? Do you know? No, you absolutely and unequivocally have no idea what white is actually being produced. You can't tell me that it's D65 because you don't know what it is, nor could you ever derive what the xy coordinates are of the color that you've just arrived at. You have no idea.

Quote:


And that's fine, but as I said in my initial post, I wasn't looking for precision at this time, just a step closer than stock settings.

You could throw a hammer at your TV and get closer than stock settings... that doesn't make a hammer a useful calibration tool...

Quote:


I think I can trust my camera more than my eyes, as I know for a fact I've gotten used to sets that were way too warm or way too cool, without even noticing it until they were fixed for me.

You should trust neither. That is your error. I am beginning to grow a little bit frustrated here with your stubbornness.

Let me say this again so it sinks in: YOU ARE NOT MEASURING ANYTHING AT ALL.

YOU ARE NOT MEASURING COLOR WHEN YOU USE A CAMERA

Say it with me: "When I point a digital camera at my TV, I am NOT measuring color."


Quote:


Most people, unless the colors are obviously off, won't see their sets looking bad initially, but will appreciate their sets much more once they have been calibrated.

True. If you want to correct that, you need to achieve accurate colors, and in order to do that you need a measurement device that is capable of measuring color and telling you what color is actually being produced. A CAMERA IS NOT CAPABLE OF DOING THIS!

Quote:


I agree, and I am certain I will continue to research this and I plan to eventually know how to calibrate my sets the right way lol but for now, at least this is good practice, and hopefully my set has slightly better color, even if it's just a small amount better than it was. Eventually I'll make it even better.

ChrisWiggles is offline  
post #27 of 208 Old 09-18-2010, 01:48 PM - Thread Starter
Advanced Member
 
morphinapg's Avatar
 
Join Date: Jun 2006
Posts: 605
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 12
Quote:
Originally Posted by ChrisWiggles View Post

Because the response of the receptors (CIE colod matching functions) in our eyes looks like this:

http://en.wikipedia.org/wiki/File:CIE1931_RGBCMF.svg

Interesting, but as for the parts that are in the negative, wouldn't it make more sense to compensate for that by raising the g and b response curves rather than raising the r curve? When you have -r, and are limited by constraints, usually you increase b and g. But then again, that's talking RGB in the video space again, but I can only talk about what I know. If you raised G and B to compensate for the dip in R, it seems that you would have almost linear RGB, which makes a lot more sense to me.

Quote:
Originally Posted by ChrisWiggles View Post

Those are arbitrary choices of signal levels that are defined by video standards, not color science. Then you have signals that are equal in amounts of R' G' and B', then video standards define that as a neutral white, and most imaging standards (but not all) define neutral white as D65. But neutral white could just as easily be something else, like D55, or 9300K, or whatever the standard dictates.

Which is why you need to calibrate the camera to the correct white point first, before using it as a tool to measure color. No, it's not measuring it in the same way the meters do, but it is measuring it. 100%/0%/0% is red, if there was any blue or green in there, it is not pure red. The same goes for white/grey which is 100%/100%/100%. If you get something like 100%/95%/100%, then that's close to grey, but does not have enough green.

Quote:
Originally Posted by ChrisWiggles View Post

In color science, CIE Illuminante E is the equal-energy illuminant, which has uniform power across the visible light spectrum. That's more the physical "neutral white" that you might be looking for, but that is not the neutral white point chosen by video standards and that is implied by values that are equal in R'G'B' values in a video source. D65 doesn't have equal energy of light across the spectrum, and looks significantly different than the E point, as you saw in the SPD chart of D65 I posted above.

No, D65 does not have equal energy across the spectrum, but once you calibrate a camera for D65, it will appear as true white, or 100%/100%/100%.

Quote:
Originally Posted by ChrisWiggles View Post

Maybe the camera sensors will give you equal amounts of R'G'B' for the E point. Or for some other arbitrary color temperature. And it will depend entirely on the SPD of whatever the camera is pointed at. So just because a camera is giving you equal amounts of R'G'B' values tell you nothing really about the color that the camera is sensing.

Except when the camera is calibrated properly, it will be close. Yes, the different SPDs will disrupt that slightly (I'll take your word on it, even if it doesn't make sense to me) but it will be close, even if it's not perfect, and close was all I was going for.

Quote:
Originally Posted by ChrisWiggles View Post

But you're not actually tweaking anything. Do you understand that? Let me ask you a question instead: when you are tweaking your R'G'B' values so they are the same off the camera, what is your target? When you get them all equal, what white have you now arrived at? Do you know? No, you absolutely and unequivocally have no idea what white is actually being produced. You can't tell me that it's D65 because you don't know what it is, nor could you ever derive what the xy coordinates are of the color that you've just arrived at. You have no idea.

I have arrived at a color which is close to D65. I don't know or currently care about the xy coordinates, but since the camera is calibrated to display D65 at equal RGB levels, I should be fairly close to D65, even if being outside the error range of a standard calibration, but that's fine for me, for now.

Quote:
Originally Posted by ChrisWiggles View Post

Let me say this again so it sinks in: YOU ARE NOT MEASURING ANYTHING AT ALL.

YOU ARE NOT MEASURING COLOR WHEN YOU USE A CAMERA

Say it with me: "When I point a digital camera at my TV, I am NOT measuring color."

It may not be measuring xy coordinates, but that's not what I'm trying to do anyway. I calibrate the camera for D65, which means D65 white should accurately be represented as level RGB. Using that, I point it at my TV and attempt to achieve as close to level RGB as I can. I understand the difference in SPD between daylight and my TV will cause a slight variation in my color, but I also believe that variation won't be big enough for the color to be worse than when I started. As long as I have gotten the color closer to calibration, I have succeeded in what I set out to do. I never planned for this to be an entirely accurate or substitution for a real calibration, but using the tools I already have, without buying further tools, I believe I have gotten closer than I was before, and I believe I have gotten closer than I would have by just using my eyes.

Quote:
Originally Posted by ChrisWiggles View Post

True. If you want to correct that, you need to achieve accurate colors, and in order to do that you need a measurement device that is capable of measuring color and telling you what color is actually being produced. A CAMERA IS NOT CAPABLE OF DOING THIS!

A camera measures color, just in RGB space (which is a real world color space) and not in xyz. I realize you're used to measuring color in xyz and find it hard to believe RGB could accurately represent color, but it can. It might not be as accurate as xyz, but it's close enough for me for the moment.

By the way, I'm still not clear on what SPD represents, other than for some reason it changes the color in a camera and not a meter used for calibration. Is it related to the brightness of the color, or what? I just don't get what it represents.
morphinapg is offline  
post #28 of 208 Old 09-18-2010, 04:20 PM
 
ChrisWiggles's Avatar
 
Join Date: Nov 2002
Location: Seattle
Posts: 20,730
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 14
Quote:
Originally Posted by morphinapg View Post

Interesting, but as for the parts that are in the negative, wouldn't it make more sense to compensate for that by raising the g and b response curves rather than raising the r curve? When you have -r, and are limited by constraints, usually you increase b and g. But then again, that's talking RGB in the video space again, but I can only talk about what I know. If you raised G and B to compensate for the dip in R, it seems that you would have almost linear RGB, which makes a lot more sense to me.

Okay, I'm sorry but now you're starting to piss me off.

You readily admit that you don't have any background or understanding of color, color science, or how SPDs of various colors react with whatever sensors or receptors respond to those SPDs.

And you've had several people with deep understanding of color and calibration state unequivocally that you CANNOT use the camera as an objective measuring device for characterizing color. Yet you're still here adamant that what you are doing is in any way more than guessing.

And now you're trying to re-define a century of color science research because it doesn't make sense to you, despite the fact that you readily admit you really have NO idea what you're talking about here.

Quote:


Which is why you need to calibrate the camera to the correct white point first, before using it as a tool to measure color.

What are you calibrating it to? Are you pointing the camera at another D65 source and aligning it to that? If so, then what happens when you point it back at the TV, with an entirely different SPD? How do you know that the camera will respond to those two D65 sources the same way? You don't know that, and it likely won't. Again, as I said before, the BETTER way and CORRECT way to use a camera is as an optical comparator, and to match the display to the D65 reference by eye first, THEN use the camera ONLY as an optical comparator on the single TV in question.

Do you understand that there are an infinite number of sources that are D65? And they can all be totally different, despite being the same color. The camera will inherently respond to those different SPDs in different ways in many cases because the camera is NOT a CIE accurate standard observer and NEVER WILL BE. That means that you could have a dozen TVs, all of them PERFECTLY calibrated to D65, and the camera could see them all as different.

So, in order to use the camera CORRECTLY, you can ONLY use it to COMPARE colors on a SINGLE display because it's overall SPD will NOT change(very much) based on luminance. So the proper process would be to look at an accurate D65 reference, and align the TV at a particular %stim to that same color of grey by eye. THEN, you point the camera at the TV, you measure how the camera responds to that particular TV and use that response to align the greyscale across the display's luminance range. Fundamental to this though, is the limitation that you CANNOT compare D65 from one display or source to another display because the camera likely will respond differently.

Quote:


No, it's not measuring it in the same way the meters do, but it is measuring it.

Measuring WHAT??

WHAT is it measuring? You have absolutely NO idea what it is measuring. All we DO know for certain is that it is NOT measuring color.

Say it with me: A CAMERA DOES NOT MEASURE COLOR.

A camera only responds to color in an arbitrary way. We cannot make ANY concrete conclusions about what the color actually WAS that the camera was responding to unless you spend a ridiculous amount of time profiling the camera AND what it is the camera is measuring in the first place. One could do that, but it requires spectrometers to do it.

Quote:


100%/0%/0% is red, if there was any blue or green in there, it is not pure red.

In the video signal, sure. But you could point a camera at a 100% red pattern, or a 100% green pattern and it is likely to return a value like 100%/3%/4% or anything it chooses. You have no idea what it will tell you.

Quote:


The same goes for white/grey which is 100%/100%/100%. If you get something like 100%/95%/100%, then that's close to grey, but does not have enough green.

Sure, but again what will the camera see? Point a camera at a calibrated D65 display, or several of them, and you will see that on every display it will give you entirely different numbers, and you would be EXTREMELY lucky if you just HAPPENED to get a display that a camera would report as being equal amounts of "RGB."

Quote:


No, D65 does not have equal energy across the spectrum, but once you calibrate a camera for D65, it will appear as true white, or 100%/100%/100%.

Only for THAT specific D65. Not for others.

Quote:


Except when the camera is calibrated properly, it will be close. Yes, the different SPDs will disrupt that slightly (I'll take your word on it, even if it doesn't make sense to me) but it will be close, even if it's not perfect, and close was all I was going for.

Thank you for at least taking my word one something.

Quote:


I have arrived at a color which is close to D65.

Have you? You have no way of knowing. And if you did the comparing by eye, you actually would have greater confidence that they were the same color because your eyes being human eyes are actually seeing color the a human does so you can be assured that you are actually quite close by visually comparing the greys. If you have a D65 reference, do it by eye, THEN use the camera to make your greyscale flat. D65 measurements with a digital camera are NOT portable to other stimuli.

Quote:


I don't know or currently care about the xy coordinates,

Then WHAT THE F**K? That's the whole point of achieving a neutral grey! Every color is mapped to xy coordinates. If you want to align greyscale accurately, you HAVE to care about the xy coordinates, otherwise you've lost your marbles completely.


Quote:


but since the camera is calibrated to display D65 at equal RGB levels, I should be fairly close to D65, even if being outside the error range of a standard calibration, but that's fine for me, for now.

Seriously: just use your eyes. All you're doing here is pretending to use a tool that isn't capable of telling you what you want it to tell you. And now you're just starting to get annoying because you don't want to accept that fact, and you don't want to put any real effort into understanding why that is.

Quote:


It may not be measuring xy coordinates, but that's not what I'm trying to do anyway.

Then what are you trying to do? Make a tuna sandwich? WTF?

Quote:


I calibrate the camera for D65, which means D65 white should accurately be represented as level RGB.

Only for THAT D65, not for others.

Quote:


Using that, I point it at my TV and attempt to achieve as close to level RGB as I can.

So what will the result be? Will it be D65? NO!

You're better pointing your EYES at the TV and comparing the D65 reference to the display and making them look the same. That way you are actually aligning the TV to D65 and not some arbitrary value which may or may not be D65 at all.

Quote:


I understand the difference in SPD between daylight and my TV will cause a slight variation in my color, but I also believe that variation won't be big enough for the color to be worse than when I started. As long as I have gotten the color closer to calibration, I have succeeded in what I set out to do. I never planned for this to be an entirely accurate or substitution for a real calibration, but using the tools I already have, without buying further tools, I believe I have gotten closer than I was before, and I believe I have gotten closer than I would have by just using my eyes.

Believe whatever you want. It's a free country. Just don't argue with people who know what they're talking about when they tell you you're wrong.

Quote:


A camera measures color, just in RGB space (which is a real world color space) and not in xyz. I realize you're used to measuring color in xyz and find it hard to believe RGB could accurately represent color, but it can. It might not be as accurate as xyz, but it's close enough for me for the moment.

NO a camera does NOT measure in CIE RGB space. A camera measures entirely arbitrary values which are denoted as R'G'B' in video and imaging standards, but are NOT CIE RGB values, and you CANNOT derive actual CIE RGB or XYZ values from them.

Quote:


By the way, I'm still not clear on what SPD represents, other than for some reason it changes the color in a camera and not a meter used for calibration. Is it related to the brightness of the color, or what? I just don't get what it represents.

It's the entire spectrum of light that comprises anything we see. It's the actual stimulus to which our eyes and cameras and color meters all respond.

http://en.wikipedia.org/wiki/Spectra...r_distribution

The following images (and more) can all be found here:
http://www.gelighting.com/na/busines...ion_curves.htm

Here's the SPD of a D65 flourescent:

http://www.gelighting.com/na/busines...ution/SP65.jpg

Here's the SPD of a daylight bulb:

http://www.gelighting.com/na/busines...n/daylight.jpg

Here's the SPD of some daylights:

http://www.gelighting.com/na/busines...r_daylight.jpg


How our eyes will react to those versus a camera could differ considerably. And the SPDs of various display can vary HUGELY depending on the light source. A phosphor, versus an HID lamp, versus an LED, etc are all VERY different, that even properly designed colorimeters (let alone far inferior cameras) run into significant problems which is why on some display you really have to be using a spectrometer.
ChrisWiggles is offline  
post #29 of 208 Old 09-18-2010, 04:39 PM
AVS Special Member
 
PlasmaPZ80U's Avatar
 
Join Date: Feb 2009
Posts: 7,140
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 63 Post(s)
Liked: 195
I'd say it's time for this thread to be locked. There's not much (if anything at all) left to explain.

Is it possible to calibrate a display using a high quality digital camera? No.
PlasmaPZ80U is offline  
post #30 of 208 Old 09-18-2010, 06:04 PM - Thread Starter
Advanced Member
 
morphinapg's Avatar
 
Join Date: Jun 2006
Posts: 605
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 12
Quote:
Originally Posted by ChrisWiggles View Post

Okay, I'm sorry but now you're starting to piss me off.

You readily admit that you don't have any background or understanding of color, color science, or how SPDs of various colors react with whatever sensors or receptors respond to those SPDs.

And you've had several people with deep understanding of color and calibration state unequivocally that you CANNOT use the camera as an objective measuring device for characterizing color. Yet you're still here adamant that what you are doing is in any way more than guessing.

And now you're trying to re-define a century of color science research because it doesn't make sense to you, despite the fact that you readily admit you really have NO idea what you're talking about here.

lol, don't get too worked up about it. I wasn't trying to challenge your knowledge on the subject. I just thought it sounded odd given what I knew about color. For example, let's say a point on that curve corresponded to -73/95/15 RGB, most LDR systems would clip the red to 0, resulting in 0/95/15, but you could also compress that red into the LDR range, resulting in 0/131/68, which refers to the same color, but obviously raised the G and B portions to compensate for the out of range R portion. I was just wondering why the R curve gets raised instead of the G and B to compensate for that negative portion, that's all.

Quote:
Originally Posted by ChrisWiggles View Post

What are you calibrating it to? Are you pointing the camera at another D65 source and aligning it to that? If so, then what happens when you point it back at the TV, with an entirely different SPD? How do you know that the camera will respond to those two D65 sources the same way? You don't know that, and it likely won't. Again, as I said before, the BETTER way and CORRECT way to use a camera is as an optical comparator, and to match the display to the D65 reference by eye first, THEN use the camera ONLY as an optical comparator on the single TV in question.

And I would do that if I could, however, I do not have a D65 light source indoors near my TV that I can use as a reference, only daylight outside, which means I can only use the camera as a reference, and nothing else unfortunately. Otherwise I would. I realize the differences between daylight's spd and the TV's SPD will cause a slight shift in the perceived color by the camera, but as long as it's in the "ballpark" (as you said it would be) than I'm okay for now. Eventually I'll narrow it down even further.

Quote:
Originally Posted by ChrisWiggles View Post

Do you understand that there are an infinite number of sources that are D65? And they can all be totally different, despite being the same color. The camera will inherently respond to those different SPDs in different ways in many cases because the camera is NOT a CIE accurate standard observer and NEVER WILL BE. That means that you could have a dozen TVs, all of them PERFECTLY calibrated to D65, and the camera could see them all as different.

Yes, I realize this, and as you said before, they will be different, but not a huge difference, meaning I'll still be in a ballpark range. However, something strikes me as odd. How can different light sources, that have different SPDs, look like the same color to us? Does that mean our eyes are inaccurate, and that maybe the camera is seeing color in a slightly more real world way, a way that our eyes can't see? If one D65 has an SPD which has more green looks the same to us, aren't we seeing it wrong?

Quote:
Originally Posted by ChrisWiggles View Post

Measuring WHAT??

WHAT is it measuring? You have absolutely NO idea what it is measuring. All we DO know for certain is that it is NOT measuring color.

Say it with me: A CAMERA DOES NOT MEASURE COLOR.

A camera only responds to color in an arbitrary way. We cannot make ANY concrete conclusions about what the color actually WAS that the camera was responding to unless you spend a ridiculous amount of time profiling the camera AND what it is the camera is measuring in the first place. One could do that, but it requires spectrometers to do it.

We can't accurately represent the xy color, no, but we can closelyrepresent the RGB color. It won't be exact, but that's okay for now.

Quote:
Originally Posted by ChrisWiggles View Post

In the video signal, sure. But you could point a camera at a 100% red pattern, or a 100% green pattern and it is likely to return a value like 100%/3%/4% or anything it chooses. You have no idea what it will tell you.

And as long as it reads a number like that, and not something like 100%/25%/64%, then it's within an acceptable range of error to me. If This gave me a worse picture than I started with, I would be able to tell, and it hasn't.

Quote:
Originally Posted by ChrisWiggles View Post

Have you? You have no way of knowing. And if you did the comparing by eye, you actually would have greater confidence that they were the same color because your eyes being human eyes are actually seeing color the a human does so you can be assured that you are actually quite close by visually comparing the greys. If you have a D65 reference, do it by eye, THEN use the camera to make your greyscale flat. D65 measurements with a digital camera are NOT portable to other stimuli.

The way I know is because it looks better than it did before. That's all I can rely on. I don't have a D65 reference indoors to use otherwise I would.

Quote:
Originally Posted by ChrisWiggles View Post

Then WHAT THE F**K? That's the whole point of achieving a neutral grey! Every color is mapped to xy coordinates. If you want to align greyscale accurately, you HAVE to care about the xy coordinates, otherwise you've lost your marbles completely.

Neutral grey can be represented in a number of ways, using the xy like you do, RGB, HSV, or a number of other color coordinates. I can use the camera to tell me whether the color looks grey or not, but I don't necessarily need to know what the exact xy coordinates are to tell me that. I'm using the camera as a comparator like you said before, comparing my TV's image with the D65 daylight source I have, and ignoring the slight shift in color the change in SPD will give me. If that shift is too large, I will be able to tell that my tv looks worse off, but it does not, it looks better.

Quote:
Originally Posted by ChrisWiggles View Post

Believe whatever you want. It's a free country. Just don't argue with people who know what they're talking about when they tell you you're wrong.

I'm not arguing with what you know, but you also don't seem to get that I'm not trying to get withing the error range of a standard calibration. I know that is impossible with a camera, but I think I can get closer to a calibration with the camera than just using my eyes, since I don't have a D65 reference to use near my TV.

Quote:
Originally Posted by ChrisWiggles View Post

It's the entire spectrum of light that comprises anything we see. It's the actual stimulus to which our eyes and cameras and color meters all respond.

http://en.wikipedia.org/wiki/Spectra...r_distribution

The following images (and more) can all be found here:
http://www.gelighting.com/na/busines...ion_curves.htm

Here's the SPD of a D65 flourescent:

http://www.gelighting.com/na/busines...ution/SP65.jpg

Here's the SPD of a daylight bulb:

http://www.gelighting.com/na/busines...n/daylight.jpg

Here's the SPD of some daylights:

http://www.gelighting.com/na/busines...r_daylight.jpg


How our eyes will react to those versus a camera could differ considerably. And the SPDs of various display can vary HUGELY depending on the light source. A phosphor, versus an HID lamp, versus an LED, etc are all VERY different, that even properly designed colorimeters (let alone far inferior cameras) run into significant problems which is why on some display you really have to be using a spectrometer.

Thank you for that information, but like I said before, why do we see different SPDs of D65 as the same color when obviously the readings show that they are different? I find that quite odd.
morphinapg is offline  
Reply Display Calibration

User Tag List

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


Forum Jump: 

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

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