Skip navigation
This discussion is archived
Currently Being Moderated

Working Without White Balance

Mar 31, 2012 11:58 AM

Camera: Nikon D90

Adobe CS5 Design Suite

 

My goal is to work in Photoshop and Camera Raw without using White Balance.

 

As I understand, the camera stores within the raw NEF file what was seen by the camera chip without correction. For example, if the camera chip sees say 10,000 red photons, this is stored as say 200. If the camera chip sees 20,000 red photons, this is stored as 400. Hence, it is linear. Of course, the actual storage is as a Bayer array.

http://www.cambridgeincolour.com/tutorials/camera-sensors.htm

In theory, from the raw NEF file, one should be able to reconstruct the image that fell upon the camera chip.

 

On the camera itself, there are many camera White Balance settings. However, it lacks a "None" i.e. "As Is" setting meaning do not assume anything about the light source and make no correctons. Just record the photon levels and give me back the same image that was focused on the camera chip.

 

As an experiment, I shot a daylight scene using 2,500K (obviously, an extreme value) for the white balance. When I look at the raw file in Camera Raw, the "As Shot" shows a heavy blue tint. Hence, Camera Raw was using the  2,500K white balance selecting of the Nikon for the "As Shot" setting in Camera Raw. Looking at the Camera Raw basic panel White Balance drop down  items, there is no "None". Hence, Camera Raw always assumes there is a White Balance and makes corrections.

 

Why would I want to do this? For example, I have multiple varied color light sources. I want to see exactly the image my eye sees. I do not want the camera making corrections since there is not a single light source for it to make assumptions about how to correct.

 

Is there a way in Photoshop or Camera Raw to show me the image without any white balance correction. Thanks in advance.

 
Replies
  • Currently Being Moderated
    Mar 31, 2012 6:44 PM   in reply to Petrula

    Petrula wrote:

     

    Is there a way in Photoshop or Camera Raw to show me the image without any white balance correction.

     

    Nope...it's impossible (and it's not useful either). There are some apps that can read the undemosiaced raw file. The first thing you notice is it's real green. Wanna know why? Cause there are 2x the green photo sites as there are red or blue. So, a raw file with zero white balance would look like crap.

     
    |
    Mark as:
  • Currently Being Moderated
    Mar 31, 2012 8:06 PM   in reply to Jeff Schewe

    The colors recorded by the camera sensor are a function of the light source color that is illuminating the scene.  The white balance input to the RAW converter is the color of the light source and this input is required in order to convert the recorded values to the correct values of the object.  If you input an incorrect value of the light source like you did when you set a 2500K value in the camera and photographed a scene in direct daylight, which is in a light source of over 5000K you get an incorrect color balance when you converted the data using the 2500K value. 

     

    What you are requesting is impossible and has no value.

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 1, 2012 11:40 PM   in reply to Petrula

    Petrula,

     

    here is one photo without whitebalancing and in raw color space

     

    3379.jpg

     

    It was demosaicked, so there are no 2x more green pixels as in original raw (mentioned by Jeff), but it's still overly green, because in most cameras, green pixels are about 2 times more sensitive to the light than red and blue. Also, since it is in raw color space, but displayed in sRGB, it is less saturated than it should be, because sensor is less sensitive to color than our eyes. As you can see, there is no much use of this image

     

    I used dcraw utility for this. It would be possible to get this even with ACR using a profile, specially made for that purpose, but it doesn't make much sense ... Camera simply doesn't see colors the same way as you

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 2, 2012 7:10 PM   in reply to Petrula

    Hi Petrula,

     

    it is shall be possible, make a backup of the original raw file, correct WB tags in raw file (use something like exiftool) to make channel multipliers equal (a-la UniWB) and that's it... certainly ACR or LR are just not the tools created for such work (you need a raw converter like RPP, it has an option to do this w/o modification to the raw file), but they will open the raw file w/ such WB... the only issue is if Nikon encodes the tags... in my example I use Panasonic raw, Panasonic being a good citizen does not encode raw channel multipliers (that is what WB is and not some kelvin/tint sliders - ignore what ACR shows in the screenshot below) :

     

    http://img9.imageshack.us/img9/6941/acruniwb.jpg

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 2, 2012 10:51 PM   in reply to deejjjaaaa

    I'm not sure that you got exactly what you wanted, because 2000/-150 is low limit of those sliders, so ACR didn't show picture as it should. But I think you were quite close. Whatever the case, it is clear that there is no use of "no wb" option

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 2, 2012 11:05 PM   in reply to Vit Novak

    > so ACR didn't show picture as it should...

     

    may be, but if ACR does not show the WB applied to the image as set in a raw file itself it is a bug, because WB = UniWB can be set using a perfectly valid method off the target... you do not like magenta targets ? how about me shooting a legit neutral WB target in a room illuminated by magentish light (through a proper gel) - shall be the same effect and a legit WB in camera ?

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 2, 2012 11:39 PM   in reply to Vit Novak

    > it is clear that there is no use of "no wb" option

     

    except that "no wb" is a perfectly legit WB where raw channel multipliers are equal

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 2, 2012 11:42 PM   in reply to deejjjaaaa

    It could be a bug or feature, depending how you define it

     

    What I suppose happened in your case is that ACR calculated temperature/tint from RGB multipliers in the raw file. Since it was out of ACR range for those sliders, it moved sliders to the range limit (it is very unlikely that ACR calculated exactly 2000/-150), so result is probably slightly different than it should be, which can be checked by dcraw where you can define rgb multipliers (like I did). But it is actually unimportant - your result generaly confirms the same as mine - it is overly green because of higher sensitivity of green pixels, so "no wb" option has no practical use for photographer

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 3, 2012 6:06 AM   in reply to Vit Novak

    > it is overly green because of higher sensitivity of green pixels, so "no wb" option has no practical use for photographer

     

    again - imagine using a filter/gel to equalize the raw channels, which is a very practical use... what will be the WB ? again, it might not be a good idea for ACR users to have an option literally saying "no WB" (because there is always some "wb") along with options like "Daylight", but ACR shall allow to use its sliders to go to the desired WB and that answers the question whether it is bug or feature... if it is a perfectly valid WB (and within a practically achievable range with retail filters/gels) and ACR can't handle it - it is a bug.

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 3, 2012 7:49 AM   in reply to deejjjaaaa

    Well, I'm not trying to defend Adobe here, as they are not paying me for that ... they just set some reasonable limits for WB sliders (like camera manufacturers - on upper models you can also set WB in K within certain limits, for instance 400D with hack has range 1800 - 11000 K) and probably didn't expect someone will be playing with raw metadata. However, there are other metadata you can use to avoid this problem if you really want, at least in dng (gain of channels, color matrices etc). So you can eventually get the same result as if you use some preset WB (for instance daylight) in the first place ...

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 3, 2012 9:42 AM   in reply to Vit Novak

    and I am not attacking you... what I mean is that "WB range" close to UniWB is a normal range and somebody (and there are people like this) who use filter (lens, outdoors) or gel (indoors over flash or lights) to saturate all channels equally (one can argue that w/ modern cameras like D7000/D800/etc it is an overkill - but there are still cameras where one more stop of exposure delivered to one or two channels makes a difference) might actually need that... but then I understand that most probably those people are not going to use ACR anyways :-)... I like the way it is done (at least UI-wise) in Iridient Rawdeveloper - you have kelvin/tint sliders for those who want and some form of channel multipliers for a few others and you can switch between 2 options to work with WB at will... peace.

     
    |
    Mark as:
  • Noel Carboni
    23,510 posts
    Dec 23, 2006
    Currently Being Moderated
    Apr 3, 2012 10:02 AM   in reply to Petrula

    Petrula, an idea:

     

    Take a photo of a calibrated gray card with known lighting, which you consider to be your reference for ''white''.  Maybe this is in noonday sunlight or a particular set of studio lights.

     

    Open that photo, sample that gray card in Camera Raw with the white balance dropper, and note the Color Temperature and Tint values.  Sample in several places to ensure noise doesn't skew the result, and average the results.

     

    Now, for other images you shoot, which you'd like to be ''absolutely colorered'' (but actually really relative to your reference for ''white'' above), just dial in those same Color Temperature and Tint values.  You could even set these up to be a preset or your Camera Raw defaults.

     

    -Noel

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 6, 2012 7:18 AM   in reply to Petrula

    Petrula wrote:

     

    Camera: Nikon D90

    Adobe CS5 Design Suite

     

    My goal is to work in Photoshop and Camera Raw without using White Balance.

     

    As I understand, the camera stores within the raw NEF file what was seen by the camera chip without correction. For example, if the camera chip sees say 10,000 red photons, this is stored as say 200. If the camera chip sees 20,000 red photons, this is stored as 400. Hence, it is linear. Of course, the actual storage is as a Bayer array. In theory, from the raw NEF file, one should be able to reconstruct the image that fell upon the camera chip.

     

    On the camera itself, there are many camera White Balance settings. However, it lacks a "None" i.e. "As Is" setting meaning do not assume anything about the light source and make no correctons. Just record the photon levels and give me back the same image that was focused on the camera chip.

     

    The sensor is linear, but the sensitivities of the red, green and blue channels differ, so one must take these differing sensitivities into account when rendering the image from the raw data. As demonstrated by others, channel multipliers of unity (UniWB) do not yield the intended result. To obtain your intended result, empirically determined multipliers would be necessary. When used for white balance, these multipliers are chosen to that a white target will have equal RGB values in the rendered image. For your purpose, multipliers would need to be chosen so that the white target would assume the color of the incident light. In other words, if the white target were illuminated with reddish light, the image would be reddish and not white.

    Why would I want to do this? For example, I have multiple varied color light sources. I want to see exactly the image my eye sees. I do not want the camera making corrections since there is not a single light source for it to make assumptions about how to correct.

     

    Is there a way in Photoshop or Camera Raw to show me the image without any white balance correction. Thanks in advance.

     

    On the camera itself, there are many camera White Balance settings. However, it lacks a "None" i.e. "As Is" setting meaning do not assume anything about the light source and make no correctons. Just record the photon levels and give me back the same image that was focused on the camera chip.

     

    As an experiment, I shot a daylight scene using 2,500K (obviously, an extreme value) for the white balance. When I look at the raw file in Camera Raw, the "As Shot" shows a heavy blue tint. Hence, Camera Raw was using the  2,500K white balance selecting of the Nikon for the "As Shot" setting in Camera Raw. Looking at the Camera Raw basic panel White Balance drop down  items, there is no "None". Hence, Camera Raw always assumes there is a White Balance and makes corrections.

     

    Why would I want to do this? For example, I have multiple varied color light sources. I want to see exactly the image my eye sees. I do not want the camera making corrections since there is not a single light source for it to make assumptions about how to correct.

     

    Is there a way in Photoshop or Camera Raw to show me the image without any white balance correction. Thanks in advance.

    I infer that you want the captured image to reflect the spectra of the actual scene luminances (or more likely a metameric tristimulus match) rather than how you perceive them. The human visual apparatus exhibits chromatic adaption, whereas the sensor does not. The CIE xyz model assumes complete chromatic adaption and the color models based on CIE models need a white point to operate, so you are attempting to use ACR in a manner other than for which it was intended. There is an interesting exercise on the Stanford web site that shows how one can use a digital camera (the Nikon D70) in conjunction with a spectrophotometer to obtain spectrophotometric data from the camera. To recreate the actual spectra, two additional colored filters are required, but if one merely needs a metameric tristimulus match, perhaps only the three native CFA filters of the camera would be needed.

     

    http://scien.stanford.edu/pages/labsite/2006/psych221/projects/06/clin tonk/index.htm

     

    Going back to ACR, experience indicates that if one wants to recreate the appearance of a colorful sunset or a candlelit scene (cases that exceed the limits of chromatic adaption), a daylight color balance works best. It would seem intuitively that an equal energy illuminant (CIE Illuminant E) would be appropriate. This corresponds to a correlated color temperature of approximately 5455K.

     

    http://en.wikipedia.org/wiki/Standard_illuminant

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 6, 2012 10:58 AM   in reply to Petrula

    No.  For multiple images with different lighting conditions, you're much better off (quality-wise) white-balancing the images individually and then combining the results. 

     
    |
    Mark as:

More Like This

  • Retrieving data ...

Bookmarked By (0)

Answers + Points = Status

  • 10 points awarded for Correct Answers
  • 5 points awarded for Helpful Answers
  • 10,000+ points
  • 1,001-10,000 points
  • 501-1,000 points
  • 5-500 points