5
\$\begingroup\$

I am capturing raw images using N900 camera. It gives me DNG files. I have some question regarding the colour temperature of these raw images.

Objective:

I am trying to map the CIE-XY Camera white point values given in this DNG image to its colour temperature value. The DNG tag is :As Shot White XY

Steps in Experiment :

  1. For this I have set the White balance mode on N900 capture application to manual mode. All other settings like Shutter speed/Aperture/Focus are in Auto mode.

  2. Then I set a particular colour temperature using a slider provided by the app.

  3. Capture raw image of a scene(general) indoors.

  4. When I inspect the metadata present in this DNG file, it has two CIE-XY white point values: As shot XY. Say, x1,y1.

  5. Then keeping the colour temperature setting constant, I just capture a different scene, by slightly moving the camera by very small amount.

  6. Now when I see the XY values, they seem to be different(slightly) than earleir image ones. Whereas I thought since camera is in manual WB mode, and colour temperature set is unchanged, it should have given same white point values.

Question:

  1. Am I correct in thinking that 'As shot White XY' values present in DNG metadata give the white points at that set colour temperature?

  2. If yes, why does it change in above setup. Does colour temperature depend upon the CMOS sensor temperature or some other parameter like exposure (Aperture/Shutter speed)?

Any pointers regarding WB, Colour temperature, CIE coordinates, XY white points would be useful.

\$\endgroup\$
3
  • \$\begingroup\$ Are there multiple light sources? The ratio of reflectance from different sources will change when shooting a different scene, and this will affect WB. \$\endgroup\$
    – Imre
    Commented May 16, 2011 at 16:38
  • \$\begingroup\$ @Imre But the camera WB setting is set manually in his test, so it shouldn't be changing in response to what's in the scene. \$\endgroup\$
    – coneslayer
    Commented May 16, 2011 at 17:11
  • 1
    \$\begingroup\$ Could you give some clarification of how much "slightly" actually is? A change of a couple percent in the color temperature of a scene is probably quite normal, as illumination, from natural or artificial sources, is rarely perfectly constant. Particularly with fluorescent tubes, artificial lights tend to oscillate with current frequency, and daylight fluctuates with changing atmospherics. \$\endgroup\$
    – jrista
    Commented May 16, 2011 at 17:54

2 Answers 2

2
\$\begingroup\$

I doubt you're going to get a real answer (this really isn't -- it's really a comment that's way too long to fit in a comment box).

The problem is fairly simple: neither the question nor the answer is really related to photography. Rather, they're both related to how the firmware of the particular camera in the particular phone you're dealing with happens to be designed, and what data it chooses to encode into the raw files it produces.

The answer would/will apply (almost?) exclusively to that particular camera/phone, not to photography in general. Just for example, when/if I set a manual white balance in my camera, and open the raw file its produces, the "as shot" white balance that's encoded in the raw file seems to remain constant across fairly radical changes in lighting. I just did a quick check of a couple pictures of the wall of my office, first under artificial light, and then opened the blind to get sunlight. The apparent color of the wall, as shown in the picture changes quite noticeably from one shot to the other, but the the "as shot" white balance stays precisely constant.

I'll add that the "as shot" white balance does seem to be coded indirectly -- in particular, although the number given for the color temperature doesn't change from one shot to the other (even though the white balance clearly has changed), different programs opening the file seem to have slightly different interpretations of how to convert what is included back to a color temperature. Just for example in the test I mentioned above, I set the balance in the camera to 5500K. When I open the files with ACR, it says they were shot at 5600K. Bibble, on the other hand, says they were shot at 5231K.

\$\endgroup\$
0
\$\begingroup\$

Depending on the camera, some of them may record the WB as what is measured by the camera's processor when it analyzes the raw image data. In other words, even though one may have a specific color temperature and WB correction selected (which the camera does apply to the image data to produce the preview image for raw files or JPEG image) the EXIF tag may be what the camera has measured to be the WB based on the contents of the raw image data.

(Please don't mistake this answer as an argument that this is acceptable. This answer is not that argument. It is only an observation that this is the case with some cameras.)

Another thing one should consider in the case stated by the question with regard to a phone camera is to ask how the selected WB is being implemented. It might not be as straightforward as one might think.

If the only variable you are allowed to specify is color temperature along the Kelvin scale, then the camera is probably still doing green ←→ magenta¹ adjustment automatically. This can alter the R,G,B multipliers between two different shots, which may be what the camera is using to compute the "WB" value recorded in the EXIF tag.

¹ The green ←→ magenta axis is the "other half" of white balance that is more or less orthogonal to the blue ←→ amber color temperature axis.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.