Ad
Ad
Ad
Pages: [1] 2 »   Bottom of Page
Print
Author Topic: neutral white balance cards/tools  (Read 8837 times)
Nino Loss
Sr. Member
****
Offline Offline

Posts: 433


« on: October 05, 2010, 01:44:17 PM »
ReplyReply

(Please forgive for reformulating here, for it's aspect pertaining to shooting gear, a question to which I got no answer in the C1 section of this forum:)

The setup:  studio tethered c1

Most image processors let you use a "white balance tool" that allows you simply pick a white/neutral reference in an image. I, like many others, click on a gray card. Depending on where on the card you click you will get very different results, even if you do as good as you can to avoid all known pitfalls associated with this technique.

Even though I continue to use them, I am not satisfied with the different X-rite gray cards, they all give me too warm and reddish results. The "digital gray Kard" is not better, and Expo-disc, to a RAW work-flow, for me, is useless. What do you do?

Is there serious comparison/test data available for gray cards? What do you use?

kind regards

nino
« Last Edit: October 05, 2010, 01:50:53 PM by Nino Loss » Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3455


« Reply #1 on: October 05, 2010, 02:06:55 PM »
ReplyReply

Depending on where on the card you click you will get very different results, even if you do as good as you can to avoid all known pitfalls associated with this technique.

Even if the card is neutral, and that's an 'if', ambient light and thus local colored objects will change the reflected colors, even across a spectrally neutral reflector.

Quote
Is there serious comparison/test data available for gray cards? What do you use?

1. WhiBal
2. Babel White target



Cheers,
Bart
Logged
Nino Loss
Sr. Member
****
Offline Offline

Posts: 433


« Reply #2 on: October 05, 2010, 02:27:13 PM »
ReplyReply

Bart, thank you for your prompt reply!

Even if the card is neutral, and that's an 'if', ambient light and thus local colored objects will change the reflected colors, even across a spectrally neutral reflector. [...]


Whatever card I use, and whether I use camera Raw or C1, as mentioned, I get very different results, mainly, and supposedly because of the reasons you bring. Only, I want to remain focused on the practical side of things. I am working in a tethered studio setup for portraits. As I charge a quite some money, my costumers become very big specialists in detecting clothing, hair and skin tone issues. And they do have good eyes. So variation in accuracy of +- 400K and +-20 in tint, is huge!

Is the only solution to this just to click around and playing in this area of +-400K and +-20 in tint until I find something satisfying? (I will later warm the picture up a little most of the time, but that's an other story)

In C1, I crop the image down to the whole gray/neutral card area (mostly the classic color checker) and apply one of the two Auto white balance possibilities. All of which give me very different results, not just in numbers, in looks. 6200K/3.1 (picker), 6044K/1.9 (Auto), 5891K/3.1 (Mode "auto"). It is mot often this last lowest number, which appears the most accurate.

For example, a picture taken in a studio painted with neutral gray, where there is nothing else except a neutral gray table, so there is nothing that the gray card could reflect, and still big differences.

Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3455


« Reply #3 on: October 05, 2010, 03:08:35 PM »
ReplyReply

In C1, I crop the image down to the whole gray/neutral card area (mostly the classic color checker) and apply one of the two Auto white balance possibilities. All of which give me very different results, not just in numbers, in looks. 6200K/3.1 (picker), 6044K/1.9 (Auto), 5891K/3.1 (Mode "auto"). It is mot often this last lowest number, which appears the most accurate.

The only characteristics that are important for a White Balance target are neutral relection across the spectrum (to warrant no metamerism and color inconsistency), and a not too glossy surface (Lambertian reflection characteristic would be optimal) to avoid picking up isolated colors i.e. from behind the camera (e.g. photographer's shirt color...). That way the color of the light falling on from all directions will be accurately integrated/reflected towards the camera.

The Camera however is not a very accurate measuring instrument because it only records 3 (partially overlapping) spectral bands and needs to reconstruct spectrally accurate colors from that. That process is particularly difficult if the light source doesn't have uniform emission in those bands. Therefore the quality of the light matters, continuous spectrum is better than spiky spectral emissions, and a good camera profile also helps.

If those conditions are met, then the White Balance picker should be able to do a good job. The 'Auto' modes use unknown logic (black magic ;-)) to arrive at a so called "correlated color temperature" from a larger scene, which can only be accurate under ideal circumstances.

Quote
For example, a picture taken in a studio painted with neutral gray, where there is nothing else except a neutral gray table, so there is nothing that the gray card could reflect, and still big differences.

The WB picker should be spot on, but not all gray cards are all that gray. The 2 mentioned above are relatively good (WhiBal), and exceptionally good (the Babelcolor White target).

Cheers,
Bart
Logged
Nino Loss
Sr. Member
****
Offline Offline

Posts: 433


« Reply #4 on: October 05, 2010, 06:12:27 PM »
ReplyReply

[...]That process is particularly difficult if the light source doesn't have uniform emission in those bands.

Therefore the quality of the light matters, continuous spectrum is better than spiky spectral emissions, and a good camera profile also helps.

That's why I do my tests with natural noon north-light falling into the studio, even though I do hold a lot of my strobes light quality.

For this I shoot RAW (with a 5d2 and a 85/1.8 ) with a profile from Phase One.

Quote

If those conditions are met, then the White Balance picker should be able to do a good job.

well it should but it doesn't. Neither in Camera Raw nor in C1. It is always too warm and too reddish.

Quote
The 'Auto' modes use unknown logic (black magic ;-)) to arrive at a so called "correlated color temperature" from a larger scene, which can only be accurate under ideal circumstances.


in C1 there are two Auto modes and they are very consistent if a large neutral reference is included in the scene. The image cropped to the bare neutral reference will yield the same result as the uncropped one, in both Auto modes. The difference is that one is always slightly warmer. The C1 Mode "auto" gives by far the best results when compared to any other method I tried (EDIT: but you have to include a large neutral reference in the scene of course). Hard to believe, I know. Even I just keep testing it time and time again.

Quote
The WB picker should be spot on, but not all gray cards are all that gray. The 2 mentioned above are relatively good (WhiBal), and exceptionally good (the Babelcolor White target).

A very important piece of information IMO, which Andrew Rodney, "the digital dog" reminds us of, quoted BTW on the BAbelcolor website, is that "... First thing, don't WB on gray! Try a non specular white. Half of all the data in a Raw linear capture is in the first stop of highlight. What converter are you using? "

But it does not help. The WB pickers are off anyway. !?
« Last Edit: October 05, 2010, 06:15:12 PM by Nino Loss » Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3455


« Reply #5 on: October 06, 2010, 04:15:57 AM »
ReplyReply

A very important piece of information IMO, which Andrew Rodney, "the digital dog" reminds us of, quoted BTW on the BAbelcolor website, is that "... First thing, don't WB on gray! Try a non specular white. Half of all the data in a Raw linear capture is in the first stop of highlight. What converter are you using? "

I use C1 most of the time, and it gives consistent white balance, not only on (non-clipped) white but also on a large range of grays (underexposed whites) below white. I've used the ultimate reference of the Babelcolor White Target and made a sequence of exposures, from underexposed to overexposed. The Click WB gave consistent neutral grays with color temperatures that didn't change more than 50 K or thereabouts, and a tint that didn't change more than 1 or 2, if at all. Overexposure can result in slightly larger deviations from what a correct exposure White Balances to, depending on the camera.

Quote
But it does not help. The WB pickers are off anyway. !?

That doesn't make sense. When you click on a uniform area, it should result in gray (R=G=B) of a certain luminosity. You mention that it results in reddish/warm tone. Do you mean it doesn't colorbalance the "picked" area to gray, or that other colors are too warm. In case of the latter, your camera profile needs tweaking and C1 allows to do that in a way by tweaking the colors and saving it as a new ICC.

Cheers,
Bart
Logged
Nino Loss
Sr. Member
****
Offline Offline

Posts: 433


« Reply #6 on: October 06, 2010, 02:19:39 PM »
ReplyReply

...
That doesn't make sense. When you click on a uniform area, it should result in gray (R=G=B) of a certain luminosity. You mention that it results in reddish/warm tone. Do you mean it doesn't colorbalance the "picked" area to gray, or that other colors are too warm. In case of the latter, your camera profile needs tweaking and C1 allows to do that in a way by tweaking the colors and saving it as a new ICC.


They do balance to a neutral R=G=B=Luminosity. In my case, most of the time, I pick on the white of the correctly exposed X-rite Gray Scale Balance Card. What I meant is that other colors are too warm and even gray and the more so the black are not R=G=B.

Why do you think that this may hint to an issue with the profile for this camera? I use the C1 generic profile for 5d2. In the past I tried to tweak the colors, but I was unable to achieve any usable profile. Could you tell me, or direct me to a source explaining how to properly do such a camera profile tweaking with C1?

regards
nino

Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3455


« Reply #7 on: October 06, 2010, 02:47:16 PM »
ReplyReply

They do balance to a neutral R=G=B=Luminosity. In my case, most of the time, I pick on the white of the correctly exposed X-rite Gray Scale Balance Card. What I meant is that other colors are too warm and even gray and the more so the black are not R=G=B.

It's hard to grasp without an actual example. When you say too warm, compared to what is that? Do you have known colors (and how reliable are they) that have a higher red component or is it the blue and green that are too low, or is the relative deviation more towards orange or yellow? Do you have an objective reference or is it more subjective? Are all colors affected or just some? Is the numerical evaluation off or is it an impression from looking at the display/print, and how well are they calibrated?

Quote
Why do you think that this may hint to an issue with the profile for this camera? I use the C1 generic profile for 5d2. In the past I tried to tweak the colors, but I was unable to achieve any usable profile. Could you tell me, or direct me to a source explaining how to properly do such a camera profile tweaking with C1?

When a known reference is so significantly off after Click WB that it can be seen as being too warm, it has to be the camera profile (unless there is a bug). One can tweak the color response by adjusting the tones in the Color Editor dialog, and there you can save as an ICC profile.

Cheers,
Bart
Logged
Nino Loss
Sr. Member
****
Offline Offline

Posts: 433


« Reply #8 on: October 06, 2010, 02:58:13 PM »
ReplyReply

It's hard to grasp without an actual example. When you say too warm, compared to what is that? Do you have known colors (and how reliable are they) that have a higher red component or is it the blue and green that are too low, or is the relative deviation more towards orange or yellow? Do you have an objective reference or is it more subjective? Are all colors affected or just some? Is the numerical evaluation off or is it an impression from looking at the display/print, and how well are they calibrated?

I did some tests just the other day. For this, I shoot inside the studio with exclusively natural noon north light (no clouds). I shoot tethered into c1. Proof profile set to output, ProPhotoRGB. Viewing on a very carefully calibrated and hooded PA241w.  Then I immediately compare with my assistances skin tone etc. to the monitor while the model is still sitting.

Quote
When a known reference is so significantly off after Click WB that it can be seen as being too warm, it has to be the camera profile (unless there is a bug). One can tweak the color response by adjusting the tones in the Color Editor dialog, and there you can save as an ICC profile.

Cheers,
Bart

I was aware of that last procedure. Just is this done properly. As I mentioned. I tried this already and the outcome was far from satisfactory.

regards
nino
Logged
Nino Loss
Sr. Member
****
Offline Offline

Posts: 433


« Reply #9 on: October 06, 2010, 03:52:04 PM »
ReplyReply

I did some tests just the other day. For this, I shoot inside the studio with exclusively natural noon north light (no clouds). I shoot tethered into c1. Proof profile set to output, ProPhotoRGB. Viewing on a very carefully calibrated and hooded PA241w.  Then I immediately compare with my assistances skin tone etc. to the monitor while the model is still sitting.

I was aware of that last procedure. Just is this done properly. As I mentioned. I tried this already and the outcome was far from satisfactory.

regards
nino

To tweak the profile, I placed a classic color checker in the mated viewing booth next to the monitor, shot RAW into C1 and adjusted with the color editor. What ever I did was a compromise. In the end I gave up because the C1 profile was by far the better compromise. Maybe I have to tweak for portraiture with a model, and would get the compromises I need for the skin tones?
Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3455


« Reply #10 on: October 06, 2010, 06:36:17 PM »
ReplyReply

To tweak the profile, I placed a classic color checker in the mated viewing booth next to the monitor, shot RAW into C1 and adjusted with the color editor. What ever I did was a compromise. In the end I gave up because the C1 profile was by far the better compromise. Maybe I have to tweak for portraiture with a model, and would get the compromises I need for the skin tones?

Assuming you WhiteBalanced not on the brightest patch of the Colorchecker's grayscale but on the second or third one, the default profile would probably be pretty decent. Yet you mentioned a too warm rendering when clicking WB. Is it better now that you had good light?

Skin is probably one of those colors that you don't want to render accurately, but rather 'pleasing' (whatever that is). Skin is hard to get accurate anyway, because of how it reflects/diffuses the light that falls on it. Different color temperatures of lighting (distributions of wavelengths) will produce different results. Since the sensor filter layers do not correspond to the eye sensitivities, it will be hard to match what we see anyway.

Cheers,
Bart
Logged
Nino Loss
Sr. Member
****
Offline Offline

Posts: 433


« Reply #11 on: October 06, 2010, 07:01:46 PM »
ReplyReply

Assuming you WhiteBalanced not on the brightest patch of the Colorchecker's grayscale but on the second or third one, the default profile would probably be pretty decent. Yet you mentioned a too warm rendering when clicking WB. Is it better now that you had good light?
it's acceptable, maybe even pleasing, but not a neutral reference.
Quote
Skin is probably one of those colors that you don't want to render accurately, but rather 'pleasing' (whatever that is).

I do not remember having used the WB result straight for portraiture, but I am looking for a more neutral reference to start from. I did not get one of those BabelColor targets yet, but now I just ordered one  Wink. Also, it gives a more accurate way of identifying the max exposure to the right, than the things I do now.

As you say, the other possible reason for the inaccuracy in the gray and black after WB, is maybe due to a poor profile, though I still don't know why. Again, it seams to me that it will be very difficult for me to beat P1 at profile making, except that they did build a "generic" profile for the 5d2, and that therefore, even a less than optimal "portrait" profile might maybe do better for that purpose? The reason why I see a slight possibility of doing better is that I will decide to make the compromises in a way that they should not affect skin tones, if such a thing is possible. I want to give it another try, despite the past unsuccessful attempts.

Quote
Skin is hard to get accurate anyway, because of how it reflects/diffuses the light that falls on it. Different color temperatures of lighting (distributions of wavelengths) will produce different results. Since the sensor filter layers do not correspond to the eye sensitivities, it will be hard to match what we see anyway.

Cheers,
Bart

regards
nino
Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3455


« Reply #12 on: October 07, 2010, 03:46:47 AM »
ReplyReply

I do not remember having used the WB result straight for portraiture, but I am looking for a more neutral reference to start from.

Yes, that's the idea. One doesn't necessarily want to neutralize ambient lighting, but one wants to have a neutral startingpoint and take it from there.

Quote
I did not get one of those BabelColor targets yet, but now I just ordered one  Wink. Also, it gives a more accurate way of identifying the max exposure to the right, than the things I do now.

Yes, that's the beauty of it. Not only does it provide the most neutral reference possible for an acceptable price, it also allows to make sure that one can expose to the right without clipping anything except specular highlights. There is nothing in a scene that will be brighter than an almost 100% reflector, only specular reflection and direct lightsources can be brighter. By searching for the clipping point in all channels, one can dial back exposure at will to accomodate for specular highlight color.

Quote
As you say, the other possible reason for the inaccuracy in the gray and black after WB, is maybe due to a poor profile, though I still don't know why.

Well, profiles are compromises. The best way to make compromises that suit your specific situation best is by making your own camera profiles, e.g. based on the MacBeth SGX-Rite Digital ColorChecker SG Card. The possibility offered by Capture One to save a tweaked profile is not optimal, but it may work good enough.

Cheers,
Bart
Logged
Nino Loss
Sr. Member
****
Offline Offline

Posts: 433


« Reply #13 on: October 13, 2010, 09:16:44 AM »
ReplyReply

I got to use the BabelColor white target today. What a difference! >450K and 1.5 (same controlled lighting, angle ...)! It feels and looks better. I wish it where a bit larger though. For large shots I'll have to figure out practical ways of using it.

Thanks a lot for that tip

kind regards
nino
Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3455


« Reply #14 on: October 13, 2010, 10:54:06 AM »
ReplyReply

I got to use the BabelColor white target today. What a difference! >450K and 1.5 (same controlled lighting, angle ...)! It feels and looks better. I wish it where a bit larger though. For large shots I'll have to figure out practical ways of using it.

Hi Nino,

Good to hear that the WB target itself was part of the puzzle. BTW, I assume you meant <450K and 1.5 . I think 450K is still a lot though, unless you are comparing the different (A)WB methods (different methods may deliver different results depending on their heuristics).

One of BabelColor White Target's benefits, besides spectral uniformity, is that it serves as an overexposure warning as well and can assist in an Expose-To-The-Right scenario. One variable less to consider. Its size is indeed not huge, but one only needs a limited number of pixels to do a Click WB.

Cheers,
Bart
Logged
Nino Loss
Sr. Member
****
Offline Offline

Posts: 433


« Reply #15 on: October 13, 2010, 03:01:28 PM »
ReplyReply

BTW, I assume you meant <450K and 1.5 . I think 450K is still a lot though, unless you are comparing the different (A)WB methods (different methods may deliver different results depending on their heuristics).

I meant the difference between the white balance picked on the white field of x-rite's large Gray Scale Balance Card and the one picked on the BabelColor white target, was up to 450K and 1.5 in tint. Which together is quite something, for our eyes here at least.

BTW we used it on the set the whole afternoon and evening, and the cover got already lost  Wink

regards

nino
Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3455


« Reply #16 on: October 13, 2010, 03:54:37 PM »
ReplyReply

I meant the difference between the white balance picked on the white field of x-rite's large Gray Scale Balance Card and the one picked on the BabelColor white target, was up to 450K and 1.5 in tint. Which together is quite something, for our eyes here at least.

Ah, I see. Apparently X-rite's isn't all that neutral ...  Cheesy

Quote
BTW we used it on the set the whole afternoon and evening, and the cover got already lost  Wink

I suppose Danny Pascale can arrange a replacement.

Cheers,
Bart
Logged
Nino Loss
Sr. Member
****
Offline Offline

Posts: 433


« Reply #17 on: October 13, 2010, 06:42:45 PM »
ReplyReply

You want the best white card available? It is the Fotowand white balancing card (size A4):
why do you say that this is the best wb card? What makes you think so?
Quote

http://www.fotowand.de/frameset.htm?CHOICE=/4687.htm

Never mind the horrible website. The products are truly superb. I have used them for many years.

interesting and a lot cheaper but according to the info (on one of the worst website I've ever seen, you are right), this card claims a reflectance of  84%. Babelcolor claims a reflectance of 99% 1%. and so on. Also, the material for the Fotowand white card is simple Polystyrene. BabelCOlor's white target is made of PTFE (Teflon) which is a standard reference.

One thing speaks for this Fotowand in addition to the low prize, the size. But than you could get any piece of left over polystyrene to do the same, cheaper and even bigger, I guess?

regards
nino
Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3455


« Reply #18 on: October 13, 2010, 07:00:38 PM »
ReplyReply

One thing speaks for this Fotowand in addition to the low prize, the size. But than you could get any piece of left over polystyrene to do the same, cheaper and even bigger, I guess?

The website didn't mention, or maybe I missed it, how neutral the "Fotowand" card is across the spectrum (i.e. resistant to Metamerism and Color Consistency errors). One can also wonder how resistant to yellowing their product is, over time.

Cheers,
Bart
Logged
Nino Loss
Sr. Member
****
Offline Offline

Posts: 433


« Reply #19 on: October 14, 2010, 07:48:41 AM »
ReplyReply

The website didn't mention, or maybe I missed it, how neutral the "Fotowand" card is across the spectrum (i.e. resistant to Metamerism and Color Consistency errors). One can also wonder how resistant to yellowing their product is, over time.

Cheers,
Bart

I used a foam cup quite some times. It, for sure can give very good results, for me at least, often better than most other devices and cards. But I would like to add once more, how impressed I am by the BabelColor white target. That's another league all together. I had no color issues in post production. One click.

BTW the cover for the target eventually turned up again. I'll have to tell everybody (including me ;-) to use the bracelet that came with it.

regards
nino
« Last Edit: October 14, 2010, 07:50:20 AM by Nino Loss » Logged
Pages: [1] 2 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad