12
\$\begingroup\$

I need a scientific raw photo for colour analysis. I shot with a reference gray card which has a defined Lab color space L=70*a=0*b=0.

I need to balance the photo with L=70 in camera raw. Is it correct to move the exposure level in order to obtain such value? Is there any option to input a defined L*a*b value in order to balance the photo accordingly?

\$\endgroup\$
6
  • 2
    \$\begingroup\$ I'm voting to close this question as off-topic because this is about using a camera as a measuring instrument, not for photography \$\endgroup\$
    – mattdm
    Dec 25, 2017 at 13:59
  • 25
    \$\begingroup\$ @mattdm: Scientific photography is still photography, and should be valid here. We aren't, or at least shouldn't, be just about artsy photography. Understanding the science behing photography is important, even if the photography is eventually for non-scientific purposes. Such questions therefore also have value to a broader audience. \$\endgroup\$ Dec 25, 2017 at 15:15
  • 4
    \$\begingroup\$ Scientific photography, sure. "I want to use a camera as a measuring instrument rather than to take photographs" is, by definition, not photography. \$\endgroup\$
    – mattdm
    Dec 25, 2017 at 16:30
  • 1
    \$\begingroup\$ Also, OP should specify problem exactly. What is the end result you want to achieve? \$\endgroup\$ Dec 25, 2017 at 20:25
  • 9
    \$\begingroup\$ @mattdm: How-cameras-work questions are on-topic, aren't they? I don't see why questions about using a camerat to measure light would be off topic for this site, even if it's not actually photography. (I think in this case the OP does want to capture an image of something, presumably in-focus and everything, so it is actually photography, but it would be probably on topic even if using the camera as a light-meter or something. Not automatically a good question: it still needs to explain enough details to be answerable. This one could use more detail about the goal.) \$\endgroup\$ Dec 26, 2017 at 2:44

2 Answers 2

11
\$\begingroup\$

This frequently comes up in photographic reproduction jobs where one is trying to closely approximate some other object such as a painting antique drawing. This cannot be done with typical photographs even when adjusting them to a specific, matching, LAB color. Regular photographs increase color saturation and tailor contrast, boosting the midranges and decreasing highlight contrast. This process is called "Output Referred" and most of it occurs with the image is captured and converted to a jpeg, or when a RAW image is processed by a RAW convertor. This is done to produce pleasing images, not accurate images.

Accurate photography and reproduction is a process called "Scene Referred" and it involves controlled illumination with light that is reasonably close to D50 (roughly afternoon daylight and close to that of a 5,000K black body). Strobes are often surprisingly close to D50.

An example of where this is done is to capture, process, and print a photo of a photo. Then, when they are placed side by side under the same lighting, they should closely match.

To photograph images "accurately" one wishes to capture an image that can be converted to exactly match the colors reflected with the only modification to shift the white point to D50's white point and adjust the overall gain so that the captured LAB in the image is the same as the measured reflectance LAB (which will be D50 based). Doing this precisely is not possible as the camera color filters would have to meet The Luther/Ives criteria as well as initially illuminating the photographed object with D50. However, one can come very close using tools with scene referred rendering capability. Far, far closer than just tweaking a regular photograph.

First, you need a RAW capture. Then it should be processed using a camera profile created for scene referred rendering. Freeware, DCamProf, can do this and you can get source/docs/executables here for making Photoshop ACR compatible camera profiles:

https://www.ludd.ltu.se/~torger/dcamprof.html

Once you have this, you just adjust the Exposure, Temp., and Tint sliders so that the color of the reference point in the photograph matches your LAB measured (or known, say from a Colorchecker card). Don't adjust any other sliders. - This is for old CS3 vintage Photoshop. see instructions at the end of this post for Photoshop CC

You may need to do conversions from whatever Colorspace you use (I suggest Adobe RGB or ProPhoto RGB). Bruce Lindbloom has a convenient calculator for doing this.

http://www.brucelindbloom.com/

After you have a scene referred photograph, you can print it retaining accurate color by selecting "Absolute Colorimetric Intent" which preserves colors. That is, if a color reads LAB (70,20,25) in Photoshop it should print the same and read close to that same LAB value. Most inkjets with 8 or more ink colors are capable of doing so quite precisely though they may require something called a custom profile.

Finally, always do a process crosscheck. Take a picture of a ColorChecker card then make your reproduction. They should match extremely closely aside from things like specular reflections if using glossy paper. Matte paper is preferable for tasks like these.

One last consideration. Avoid papers that have optical brighteners. These are used to make paper's look whiter by converting uV into bluish light. The problem is that you get considerable color shift depending on how much uV is in the light. It also tends to yellow within a relatively small number of years.

Added: Process for scene referred imaging in Photoshop CC with ACR using a camera DCP "linear" profile:

ACR has changed over the years. For Photoshop CC the some ACR settings need to be adjusted per "madmanchan" on the Adobe Forums. He's an Adobe dev employee and general expert on color management.

For scene referred image processing, select the "Version 1 (2003) or Version 2 (2010) workflow available under the "camera" icon and the DCP profile. Then, in the main tab (left most "iris" icon) set the "Brightness" at 50 and the "Contrast" at 25. One can also move the "Black" in slightly to mitigate glare but it should be a fairly tiny amount. Other sliders at 0.

Adobe's processing is, to say the least, opaque, but this has produced good results for me allowing imaging of a colorchecker with an average dE2000 of 2 and max of 5. It's about 3 or 4 times better than just setting the white point and the L*50 patch.

Beyond this, the program "dcraw" has specific options for linear processing.

\$\endgroup\$
5
  • \$\begingroup\$ Excellent answer. One small note though. I don't believe it's possible to correctly modify exposure in ACR, because nobody knows exactly what the exposure slider does in ACR, and it's not a simple linear scaling. See my own question about this and my own answer. Personally If I had to do this I'd write a small program. \$\endgroup\$ Dec 26, 2017 at 10:40
  • \$\begingroup\$ @AramHăvărneanu there are other raw convertors which have proper exposure control. Darktable, Rawtherapee both have it. \$\endgroup\$ Dec 26, 2017 at 17:45
  • \$\begingroup\$ @Aram The "Exposure" slider operates linearly but only if the DCP (camera profile) tone curve is also linear (scene referred). Normally DCPs do not use a linear tone curve and the exposure slider will operate non-linearly and compress highlights. A good way to check tone curve response overall is to convert a scene referred image to D50XYZ using the matrix profile that can be found at www.color.org. These are in gamma=1. Then just inspect them against the Colorchecker XYZ reflectance specs. \$\endgroup\$
    – doug
    Dec 26, 2017 at 19:51
  • \$\begingroup\$ @doug excellent information! Much appreciated. Maybe you can add it as an answer to my question? \$\endgroup\$ Dec 27, 2017 at 11:47
  • \$\begingroup\$ For some reason the brucelindbloom.com is no longer working. Perhaps its been overloaded \$\endgroup\$
    – joojaa
    Dec 27, 2017 at 12:54
20
\$\begingroup\$

Sounds a bit iffy to me... A card, by itself, DOESN'T have a fixed L*a*b* colour - that's a product of the reflectivity of the card at different wavelengths and the intensity and wavelength distribution of the illumination. In the dark, L* will be 0. Light it with a coloured light and your a*b* will change. Light it with a "white" light with a different wavelength distribution and your a*b* may change due to metamerism. Light it with a different colour temperature and things will change. Change the brightness of the illumination or the distance or angle of the card relative to the light source and your L* will change.

However, assuming you're taking the picture under the correct standard lighting that your original card L*a*b* values were measured under (Which may well be different from natural light or camera flash - commercial printing often uses D50), And assuming that what you're interested in falls in the linear portion of the sensor response in all channels (anti blooming sensors start discarding some of the charge as the pixel well fills up), then you can probably get away with tweaking the exposure level to get the card to match your desired L* values. But be aware that even if you're using standard lighting, your L* value will vary based on distance and angle to the light source - and both brightness and colour may vary due to additional light reflected onto that part of the scene from surrounding objects.

But the usual way of measuring colours when accuracy is important is to use something called a spectrophotometer - these are usually contact (or near contact) instruments that measure the reflectivity of the target patch across a whole range of wavelengths, which also allows them to calculate accurate XYZ colour values, rather than trying to work out an accurate colour from RGB filtered samples alone, which don't always get it right (as anyone who has tried photographing bluebells will probably have noticed). This is why professionally generated ICC colour profiles normally use a spectrophotometer to meaure printed colour patches.

Which isn't to say you can't do useful things from a photo - you just have to be aware of the issues and limitations.

\$\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.