Ad
Ad
Ad
Pages: « 1 [2] 3 »   Bottom of Page
Print
Author Topic: white balance cards (raw)  (Read 8487 times)
bjanes
Sr. Member
****
Offline Offline

Posts: 2718



« Reply #20 on: January 20, 2010, 06:32:30 AM »
ReplyReply

Quote from: ErikKaffehr
Another issue is exposure. If you expose for a white card it would be gray, Would you expose for a black card it would be gray. I'd suggest that using

- A whibal (which has both white and dark parts)

- Any good white card on gray background

- Or a Xrite color checker would be optimal
Eric, your comment on exposure is a good one. If you meter off a card that fills the frame, then the metered exposure will be midgray regardless of the reflectance of the card. On the D3 the result is at 12% saturation. However, when you perform a preset WB on the D3, the camera gives extra exposure to bring up the saturation to a more useful level. The Nikon manual states that one can use either a white or gray card. The Nikon guru Thom Hogan reports that in practice he gets better results with a gray target rather than a white one. A piece of white paper is suboptimal, since it might have an florescent brightener. In a pinch, a paper towel would be a better choice.

If you include a neutral card in a scene that is properly exposed, then a gray card will receive less exposure than a white card. However, the white card would exhibit a color shift if there is clipping, since the green channel blows first with daylight and most other commonly used illuminants. A WhiBal or similar card would avoid clipping under these conditions. One can then white balance in the raw converter using the neutral card reading.
Logged
BartvanderWolf
Sr. Member
****
Online Online

Posts: 3019


« Reply #21 on: January 20, 2010, 04:26:29 PM »
ReplyReply

Quote from: NikoJorj
OK, I didn't thought of high ISOs at first but that sound quite clear now that aiming the WB eyedropper at noise will just send everything overbord.

Only if single pixels are sampled, so getting a sample of a larger region will average the noise somewhat.

Cheers,
Bart
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2718



« Reply #22 on: January 20, 2010, 08:05:20 PM »
ReplyReply

Quote from: BartvanderWolf
Only if single pixels are sampled, so getting a sample of a larger region will average the noise somewhat.
Bart,

An excellent point. In Adobe Camera Raw the color sampler covers a 5x5 pixel area, and if the WB tool samples an equivalent area, one would be sampling 25 pixels. The standard deviation I gave above was for individual pixels and the standard error for the mean of 25 pixels would be s/sqrt(25). Nikon Capture NX allows one to sample an even larger area. Therefore, precision would only be a problem at very high ISOs with the D3, but with a P&S camera one could have a problem at lower ISO's (if the P&S offered raw).

Another reason for sampling the upper tonal range might be that the WB correction is often not entirely linear. A color cast in the whites would be more apparent than a cast in the shadows.
Logged
BartvanderWolf
Sr. Member
****
Online Online

Posts: 3019


« Reply #23 on: January 21, 2010, 08:50:12 AM »
ReplyReply

Quote from: bjanes
Another reason for sampling the upper tonal range might be that the WB correction is often not entirely linear. A color cast in the whites would be more apparent than a cast in the shadows.

Indeed, the (valid) technical reasons aside, perception is important.

IMHO the WhiBal satisfies the requirements for White balancing, even under harsh environmental conditions (e.g. aquarium photography). When included in the scene, it will be brighter than average, yet reduce the risk of (partial channel) clipping, and it has a built-in warning for specular reflections, and a warning for (non-specular reflection) clipping. The different sizes are useful for different shooting scenarios.

The Babelcolor White target also allows to maximize (unclipped) exposure, especially in a studio environment , small for when on the road, large for the bigger studio setups. On my 1Ds2 the different channels clipped at different levels, which resulted in an inaccurate top 1/3rd exposure stop color reproduction. That was easy to confirm reliably with the Babelcolor target, because of its huge spectral uniformity (no risk of metamerism or color inconsistency creeping into the whitebalancing).

Cheers,
Bart
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2718



« Reply #24 on: January 21, 2010, 10:37:29 AM »
ReplyReply

Quote from: BartvanderWolf
IMHO the WhiBal satisfies the requirements for White balancing, even under harsh environmental conditions (e.g. aquarium photography). When included in the scene, it will be brighter than average, yet reduce the risk of (partial channel) clipping, and it has a built-in warning for specular reflections, and a warning for (non-specular reflection) clipping. The different sizes are useful for different shooting scenarios.
Bart,
By partial channel clipping, I surmise that this happens when the right edge of the bell shaped curve of the histogram of a flat field reaches saturation. Here is a flat field of the green1 channel of my Nikon D3 at ISO 200 approaching saturation (the curve is not exactly bell shaped because of pattern noise). The mean pixel value is 16021, which would correspond to an 8 bit gamma 2.2 value of 252. The minimum pixel value of 15670 corresponds to a pixel value of 251. To guard against partial clipping, one should keep the white balance white point in the low 240s in for an 8 bit gamma 2.2 rendering. Is this what you mean?

[attachment=19604:Histogram.gif]
Logged
BartvanderWolf
Sr. Member
****
Online Online

Posts: 3019


« Reply #25 on: January 21, 2010, 01:14:20 PM »
ReplyReply

Quote from: bjanes
Bart,
By partial channel clipping, I surmise that this happens when the right edge of the bell shaped curve of the histogram of a flat field reaches saturation. Here is a flat field of the green1 channel of my Nikon D3 at ISO 200 approaching saturation (the curve is not exactly bell shaped because of pattern noise). The mean pixel value is 16021, which would correspond to an 8 bit gamma 2.2 value of 252. The minimum pixel value of 15670 corresponds to a pixel value of 251. To guard against partial clipping, one should keep the white balance white point in the low 240s in for an 8 bit gamma 2.2 rendering. Is this what you mean?

Hi Bill,

Yes, it can be caused by the righthand side of the tail being clipped (with more than 1 pixel), or by unbalanced lighting (causing one channel to clip, using one pixel), but also by a Raw converter. My 1Ds2 got unusably funky at the highest 1/3rd stop clipping point on an earlier version of Capture One, not so on ACR (but ACR sucked for other reasons), probably because it used a more reasonable value assumption for the clipped channel.

Here's an old example of 2 images, one exposed correctly and one overexposed with highlight recovery on:
Adobe CameraRaw V4.3.1:


And on CaptureOne V4.0:


Imagine what would happen if the White Balance had been taken from the 'overexposed' CaptureOne conversion ...

A Babelcolor target included in a scene at the intended exposure + bracketed shots around it would easily show when to back-off, and which exposure level to use/convert, even with a converter that doesn't do it correctly. Of course such an experiment is best done with a studio setup (all variables constant, except exposure).

With the current versions I can use just about any exposure level below 254 8-bit equivalent in all of the R/G/B channels to Whitebalance in a Raw converter and all give the same colortemperature and tint result. It's an easy Raw converter test, just shoot a sequence around clipping, click the Whtebalance, and see when it deviates from the expected values.

The WhiBal usually won't get you as near to the risky levels, so it's much easier to use for White Balancing.

Cheers,
Bart
« Last Edit: January 21, 2010, 01:42:27 PM by BartvanderWolf » Logged
Tim Lookingbill
Sr. Member
****
Offline Offline

Posts: 1055



WWW
« Reply #26 on: January 21, 2010, 03:04:43 PM »
ReplyReply

I use the WhiBal card and get decent AutoWB results from my Pentax K100D DSLR shooting Raw and processing in ACR often times I don't even have to click for R=G=B (See the sample below showing how well my camera responds to white balancing outdoor shots.) If I did click on the target it would only slightly make it bluer. Wide shots like those pictured do well but close ups usually require a custom WB using the card or else I get a reddish orange cast in this type of light.

Full spectrum light makes a big difference but I've found that camera models differ in how their AutoWB meters for any neutrals including targets with some so off they can influence the color table of an image.

This Adobe Camera Raw thread:

http://forums.adobe.com/thread/516436?tstart=90

posted having trouble with AutoWB using professionally made neutral target.

To give you an idea of the spectral reflectance of different white objects lit by a close to spectrally flat light source like the Solux I've included a shot of the WhiBal card, a Melitta coffee filter, my white painted walls and Laser printer paper that contain optical brightners. The image was edited to show exactly what my eyes saw. The default ACR settings couldn't bring out the differences without tweaks to the color temp and HSL panel.

[attachment=19610:PentaxK100DWBstudy.jpg][attachment=19611:SoluxOBwhiteLL.jpg]
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2718



« Reply #27 on: January 21, 2010, 03:44:59 PM »
ReplyReply

Quote from: BartvanderWolf
Hi Bill,

Here's an old example of 2 images, one exposed correctly and one overexposed with highlight recovery on:
Adobe CameraRaw V4.3.1:


And on CaptureOne V4.0:


The WhiBal usually won't get you as near to the risky levels, so it's much easier to use for White Balancing.
That magenta highlight recovery in Capture 1 is awful. At least ACR recovers to neutral tones, but the properly exposed shot shows green highlights. Guillermo Luijk posted an example of highlight recovery on Caucasian flesh tones where ACR highlight recovery restored blown areas to neutral whereas DCRaw used a more intelligent recovery process that looked more natural. BTW, aside from this problem, what is wrong with ACR? It is my preferred raw converter for my Nikons. I don't have Capture 1.

Bill
Logged
Guillermo Luijk
Sr. Member
****
Offline Offline

Posts: 1273



WWW
« Reply #28 on: January 21, 2010, 04:11:37 PM »
ReplyReply

I think that this magenta cast is due to the consideration by Capture One of a wrong RAW saturation point for that camera. It is not a problem about its inner algorithms nor an intent to 'predict' colour in the clipped areas.

It appears magenta when white balancing areas that are not recognized by the RAW developer as neutral, which should be the right consideration in RAW clipped areas. If the RAW developer is not aware that a clipped area must remain neutral in the final image, white balance will produce a lack of green with respect to red and blue and that means magenta. This typically happens when the considered saturation point is higher than the real one. Using a lower than correct sat point will usually end just in some information being lost, but no magenta cast. It all depends on the RAW developer implementation of course.

For instance many Canon 7D users started to report magenta highlights with their ACR/LR versions. Probably Adobe already fixed that.

DCRAW allows to practice with correct/wrong saturation points and see the consequences in the developed image:

Canon 40D's RAW developed with wrong saturation point (16224, DCRAW's default):


Correct saturation point for Canon 40D (RAW histogram):


The same Canon 40D's RAW now developed with the correct saturation point (13824):



Any area of a RAW file where a single channel got clipped cannot be used for white balancing, and any RAW developer or camera should detect this.

Regards
« Last Edit: January 21, 2010, 04:20:44 PM by Guillermo Luijk » Logged

Jonathan Wienke
Sr. Member
****
Offline Offline

Posts: 5759



WWW
« Reply #29 on: January 21, 2010, 05:13:32 PM »
ReplyReply

Quote from: Guillermo Luijk
Any area of a RAW file where a single channel got clipped cannot be used for white balancing, and any RAW developer or camera should detect this.

ACR has been "that smart" since CS2; if I try to WB on an area that has any clipping ACR will tell me to WB somewhere else.
Logged

bjanes
Sr. Member
****
Offline Offline

Posts: 2718



« Reply #30 on: January 21, 2010, 09:08:43 PM »
ReplyReply

Quote from: Jonathan Wienke
ACR has been "that smart" since CS2; if I try to WB on an area that has any clipping ACR will tell me to WB somewhere else.
Yes, but we are talking about two issues. The first is white balance and the second is highlight recovery. To test the behavior of WB with ACR and the Nikon D3, I exposed a Stouffer stepwedge at ISO 1600 and the results shown below in Rawnalize. Step 1 is completely blown in the green channel and step 2 contains a few blown pixels at the right edge of the histogram. Rawnalize gives a mean green pixel value of 252.6 with a range of 242-255. Step 3 is intact and has a green aRGB value of 232.8 with a range of 213-239.

[attachment=19626:RawnalizeClipping.png]

Here is the ACR rendering into aRGB with a linear tone curve and white balance on Step 3 which is measured on sampler 3. ACR would not allow WB on step 2. Step 2 has an aRGB value of 250, which is close to the Rawnalize value of 252.6.

[attachment=19627:WB3.PNG]

And here are the ACR screen captures white balance on darker steps as shown in the composite below. Step 10 (marker 5) is close to midgray. The RGB values are virtually identical, showing one can white balance on midgray at ISO 1600 with this camera (which has relatively good noise characteristics). WB on darker tones gives poorer results.

[attachment=19628:Composite.png]
Logged
Jack Varney
Sr. Member
****
Offline Offline

Posts: 386


WWW
« Reply #31 on: January 22, 2010, 05:24:43 PM »
ReplyReply

I have been using a Whibal for several years. Recently I read, at the Whibal site, that it is best to use the lighter grey card (using the multi card set from Whibal) for white balancing. Having used the white card for several years I have found that the light grey card gives the same kelvin value as does the white card. That is a good thing in the event you over expose your test image.

It seems to me that any target that is truly neutral (i.e. r= g=  where there are sufficient numbers (i.e. small enough steps in values) should suffice to balance the colors to provide an accurate result.
Logged

Jack Varney
adam z
Full Member
***
Offline Offline

Posts: 130


« Reply #32 on: January 22, 2010, 08:31:03 PM »
ReplyReply

Are there any strong opinions about which is better (more accurate, and/or more versatile) - the babelcolour white target, or the whibal card? Has anyone tested them side by side?

Would it be worth having both for differing situations. My concern with a flat white card is that if you shoot it front on, it would probably give a different reading than if it was angled towards the main light or into the shadow side. does it pretty much even out? Sorry if it seems like a stupid question, but I have seen the difference between the shadow and highlights when clicking wb eyedropper in LR on folds of a white T-shirt for example.

Lastly, does anyone know where I can order one in Australia, rather than having to buy from overseas.

Thanks
Logged
BartvanderWolf
Sr. Member
****
Online Online

Posts: 3019


« Reply #33 on: January 23, 2010, 03:42:12 AM »
ReplyReply

Quote from: adam z
Are there any strong opinions about which is better (more accurate, and/or more versatile) - the babelcolour white target, or the whibal card? Has anyone tested them side by side?

I use both. I don't have a strong opinion on either of them other than the facts. The Babelcolor white target is (slightly) more accurate over a wider spectral range. It is not very large, just a small pill box size, so you'd typically use it to shoot relatively close compared to the larger sizes WhiBal cards. The WhiBals are thinner, so they're easier to stow in a pocket of a bag. Both are superior in spectral neutrality compared to most other so-called White balance tools. The Babelcolor target is extremely accurate, so it can also be used to do critical testing near the clipping point (when the Babelcolor target starts to clip one or more channels, you are probably close to overexposing white objects (with typical 95% reflection) in the scene, and specular highlights will be clipped).

Quote
Would it be worth having both for differing situations. My concern with a flat white card is that if you shoot it front on, it would probably give a different reading than if it was angled towards the main light or into the shadow side. does it pretty much even out? Sorry if it seems like a stupid question, but I have seen the difference between the shadow and highlights when clicking wb eyedropper in LR on folds of a white T-shirt for example.

Neither of these targets are perfect Lambertian diffusers, so angle matters, although they are rather diffuse reflectors. I generally try to position them just like an incident lightmeter, close to the subject, surface normal aimed at the camera. All angled objects will assume a slightly different colorbalance, it's just a fact of life. It is a fact of life that you can exploit of course, e.g. by angling up a bit you'll get a bit more of the blue sky weighted in, resulting in a warmer end result after click whitebalancing. Or if you want to make sure that the green reflection of the grass is eliminated from the skintone in a portrait, you can angle it a bit down. The differences can be subtle, but they're definitively something to consider, regardless of target. Shadows are usually blue, outdoors, not much you can do about it other than using gold colored reflectors.

Quote
Lastly, does anyone know where I can order one in Australia, rather than having to buy from overseas.

AFAIK, they don't have official distributors (although there may be resellers somewhere), but they are simple to deal with over the internet. Both are sold by reputable folks, and they're very helpful.

Cheers,
Bart
« Last Edit: January 23, 2010, 03:47:40 AM by BartvanderWolf » Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 8091



WWW
« Reply #34 on: January 23, 2010, 09:31:36 AM »
ReplyReply

Quote from: adam z
Are there any strong opinions about which is better (more accurate, and/or more versatile) - the babelcolour white target, or the whibal card? Has anyone tested them side by side?

Well I can tell you the Bablecolor spec’s on their site, and measurements I’ve made of two units sync up. The spec’s are impressive. When rubber hits the road, I suspect both products will deliver the goods.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Tim Lookingbill
Sr. Member
****
Offline Offline

Posts: 1055



WWW
« Reply #35 on: January 23, 2010, 06:59:50 PM »
ReplyReply

I wouldn't believe it would matter which brand target to use. Even the Melitta #2 coffee filter works pretty good in a pinch which was shown in the image I posted above. The real question is how and when would you use it given every camera's perception of WB seems to respond differently to various light sources. When would it be a good idea to set an incamera custom WB using these targets or do it in post?

These targets would work well in event photography like at a basketball game or wedding under oddball or mixed lighting. You'ld want to set your camera to a daylight, fluorescent or tungsten preset to prevent AutoWB from responding differently to every change to angle of lens to light source variances. You'ld take one shot within your series having the target and in processing click for R=G=B, copy those settings and apply to the rest in the series.

That's all these targets are really good for because most camera's deliver decent AutoWB or Flash preset under controlled studio lighting where a small tweak by eye to one can be applied to the rest in a series. Some cameras will do good in this setup and some won't. You'll have to test.

I shoot exclusively outdoor landscapes and find I don't use the target that much because I like the way my camera's AutoWB responds to most of my outdoor scenes.
Logged
BartvanderWolf
Sr. Member
****
Online Online

Posts: 3019


« Reply #36 on: January 23, 2010, 07:59:45 PM »
ReplyReply

Quote from: tlooknbill
I wouldn't believe it would matter which brand target to use. Even the Melitta #2 coffee filter works pretty good in a pinch which was shown in the image I posted above. The real question is how and when would you use it given every camera's perception of WB seems to respond differently to various light sources. When would it be a good idea to set an incamera custom WB using these targets or do it in post?

In-camera white balance only (and I don't mean it in a derogatory manner) matters to the JPEG preview embedded in the Raw data  file.

Quote
These targets would work well in event photography like at a basketball game or wedding under oddball or mixed lighting. You'ld want to set your camera to a daylight, fluorescent or tungsten preset to prevent AutoWB from responding differently to every change to angle of lens to light source variances. You'ld take one shot within your series having the target and in processing click for R=G=B, copy those settings and apply to the rest in the series.

Sure, I agree. But not everybody is shooting event photography, and in-camera WB is only catering to the in-camera White Balance... However, when you need to address the actual in-scene WB, then you'd want to be able and do it without hesitation with regards to metamerism and/or color inconsistency.

Quote
I shoot exclusively outdoor landscapes and find I don't use the target that much because I like the way my camera's AutoWB responds to most of my outdoor scenes.

Maybe, but I doubt that your camera's AWB (or any camera's for that matter) will be able to take out the green cast of outdoor landscapes (which tends to overcorrect to magenta) effectively. I certainly perfer to have a solid reference for my landscape images, without having to second guess the actual WB offset towards accurate color (as a basis to achieve preferred color).

Cheers,
Bart
Logged
Tim Lookingbill
Sr. Member
****
Offline Offline

Posts: 1055



WWW
« Reply #37 on: January 24, 2010, 08:03:32 PM »
ReplyReply

Quote
In-camera white balance only (and I don't mean it in a derogatory manner) matters to the JPEG preview embedded in the Raw data file.
Sorry I wasn't specific enough. Incamera WB (for my Pentax K100D) gets me closer to what I want shooting Raw so I don't have to click for R=G=B most of the time. If I change incamera WB, my Raw preview's WB changes as well thus setting WB incamera DOES affect MY jpegs and Raw captures. Don't know about your situation.

If you look at the four up image containing outdoor scenes I posted previously in post #27, you'll see what I mean. Other camera's like the Canon 50D in the Adobe Camera Raw discussion link I provided shows the user has to always make a custom WB even shooting in 2PM broad daylight because his gray target shows up in ACR's "As Shot" color temp a purplish blue when shooting using his camera's AutoWB. My Pentax's AutoWB doesn't render my WhiBaL gray target that way. It's more neutral.

Quote
Maybe, but I doubt that your camera's AWB (or any camera's for that matter) will be able to take out the green cast of outdoor landscapes (which tends to overcorrect to magenta) effectively. I certainly perfer to have a solid reference for my landscape images, without having to second guess the actual WB offset towards accurate color (as a basis to achieve preferred color).
I don't get a green cast in my landscapes. Mine usually errors on the warmish side. From what I gathered reading responses from program engineers in several discussions over at ACR forum, Pentax shares more internal data with Adobe in allowing ACR to better interpret my camera's WB than Canon and Nikon. At least that's what they said when my model came out and that was about 4 years ago. Things may have changed with newer Canon and Nikon camera's especially with the new ACR camera profiles.
« Last Edit: January 24, 2010, 08:07:34 PM by tlooknbill » Logged
madmanchan
Sr. Member
****
Offline Offline

Posts: 2100


« Reply #38 on: January 25, 2010, 04:02:12 PM »
ReplyReply

Camera Raw and Lightroom are not limited to using a 5x5 pixel area for doing click-white-balance (i.e., the eyedropper tool). The actual number of pixels sampled depends on your zoom level. If you wish to use more pixels in your WB sampling, you should zoom out (e.g., to 50% or 25%, or less). If you wish to use fewer pixels in your WB sampling, you should zoom in (e.g., to 100%).
Logged

Schewe
Sr. Member
****
Offline Offline

Posts: 5256


WWW
« Reply #39 on: January 25, 2010, 04:21:20 PM »
ReplyReply

ACR and LR use 5x5 "display pixels" for the white balance at 100% zoom and below (as Eric correctly says)...but I'm pretty sure you are locked to 5x5 "image pixels" above 100% zoom so zooming in past 100% does not increase the accuracy (please, correct me if I'm wrong Eric).

Logged
Pages: « 1 [2] 3 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad