29
\$\begingroup\$

Clearly to achieve reliable results, it is essential to calibrate key parts of the imaging pipeline - particularly the computer display on which images are edited.

Few seem to question the presumption that the only really correct way of calibrating a display is with a dedicated hardware colorimeter and associated (generally proprietary) software. I wonder what kinds of results can be achieved through somewhat alternate means; in particular, use of a (profiled) camera in place of dedicated colorimeter.

There is an open source library called ArgyllCMS which is primarily a free colour management tool designed to be used with dedicated hardware colorimeters; they have collected notes on such instruments and use with their software. *

There must be some reason why they don't even bother to mention the possibility of using a camera?

http://hoech.net/dispcalGUI/ is a cross platform GUI front end for ArgyllCMS, which looks quite promising for linux based workflows or as an alternative to being up-sold to top end versions of products just because one wants to calibrate two screens, for example (although I appreciate the cost of software development, I don't really think colour-management is something particularly complicated or requiring of original thought in developers).

I can see that a device which attaches directly to the screen is going to be able to eliminate the effects of ambient light on the readings and also that some devices have facility for monitoring ambient light and compensating appropriately. Clearly the sensor in a dedicated colorimeter will have different characteristics to a camera; what are the salient features?

I've got an inkling that if results good enough to be acceptable to somewhat serious amateurs were achievable without extra specialised hardware, then it would be more widely known and discussed; however I am also generally cynical about orthodox opinions stating 'you must have x or you'll never be a proper y' - something that I'm afraid internet photography forums in particular have a tendency to epitomise.

Hopefully some clear light can be shed on the matter, I think it could be generally educative to get some more views on the subject here.

A discussion from 2005 on ArgyllCMS mailing list about use of a digital camera for printer calibration

*One thing that is notable here is that Datacolor's business model seems to be particularly oriented around not only differentiating their own Spyder products on the basis of the functionality of provided software, but also obstructing the use of the hardware by third party software.

\$\endgroup\$
4
  • 2
    \$\begingroup\$ Here is what you were searching for: Monitor Color Calibration for free using your DSLR \$\endgroup\$
    – erik
    Mar 5, 2013 at 6:38
  • \$\begingroup\$ Erik; that looks very promising. If you post it as an answer, I'll mark it as accepted (although I'm a little busy to actually check it out at the moment, hopefully it works well. I believe it should in theory, at least better than my utter-rubbish-capitalist-conspiracy colorimeter). \$\endgroup\$
    – PeterT
    Mar 5, 2013 at 16:45
  • 2
    \$\begingroup\$ What's stopping you from taking a manual white balance from a defocused image of your computer monitor displaying pure white (no ambient light), then photographing a perfectly-lit grey card with the custom white balance set (NB: colour deviation will be reversed; i.e. screen too cool will produce an image too warm), then open the resulting image in an image viewer, make a note of the specific colour channel deviations, then neutralise them by adjusting the monitor's colour controls? Not scientific, but almost. \$\endgroup\$
    – HamishKL
    Apr 15, 2016 at 11:06
  • \$\begingroup\$ Sounds more scientific than buying some product because some people on forums say you need it :) \$\endgroup\$
    – PeterT
    Apr 16, 2016 at 6:46

9 Answers 9

12
\$\begingroup\$

The obvious problem (or an obvious problem, anyway) is that relatively few cameras have (even very close to) the degree of accuracy and repeatability of exposures to make it work at all well.

Back when magazine review budgets allowed it, some of them included graphs of camera shutter speed accuracy. Especially in faster shutter speeds, it was pretty routine to see an inaccuracy of 30% or more. Apertures varied a little less, but still quite a bit. Cameras using electronic shutters probably get the exposure time more accurate, but still have the same problem with aperture.

Just for example, almost anybody who's stitched a few panoramas learns very quickly that it's routine to have easily visible "seams" where the separate pictures are stitched together. Some of this may be due to changes in ambient light, but I'd guess most are due to the camera's exposure variation. You could reduce the variation at the same exposure by averaging several shots together, but this wouldn't tell you much about how much the exposure had really changed when you did what was supposed to be (say) a one-stop change in exposure.

Bottom line: I'd be very surprised if you could get even close to the exposure accuracy needed to produce a meaningful profile. I'd bet that a reasonably careful eyeball-based profile (e.g., using Adobe's color widget) would be more accurate than what you could plan on getting from most cameras.

\$\endgroup\$
5
  • \$\begingroup\$ Interesting, thanks. I wonder how hard shutter speed and aperture accuracy would be to test independently, and why these are less routinely covered in current reviews. \$\endgroup\$
    – PeterT
    Apr 10, 2011 at 21:45
  • 6
    \$\begingroup\$ +1 For introducing the point about exposure accuracy. However, it doesn't immediately rule out @PeterT's idea. Lack of accuracy is not a problem--but poor reproducibility is. E.g., if your camera consistently takes a 1/40 sec. exposure for a nominal 1/30 sec setting, it's inaccurate but it's precise. One can measure precision by taking pictures of the same target under the same conditions at the same settings, again and again. So why not photograph a standard target on screen (no ambient illumination)? Comparing different shots will tell us how well the settings were repeated. \$\endgroup\$
    – whuber
    Apr 10, 2011 at 22:21
  • \$\begingroup\$ @whuber: I also mentioned repeatability... \$\endgroup\$ Apr 11, 2011 at 1:09
  • 4
    \$\begingroup\$ Yes, you did. But you focused on accuracy. There is some hope: one person did accurate measurements of a crummy camera shutter and found a standard deviation of 0.08 stop (circlemud.org/~jelson/lomography_shutter) even though it was over 1/2 stop too rapid. Others report that precision is a challenge at the fastest speeds but perhaps not so much at the slower ones, which makes sense mechanically (nemeng.com/leica/006b.shtml). Thus it seems possible, even likely, that a decent SLR will have highly repeatable exposure at slow to moderate shutter speeds. \$\endgroup\$
    – whuber
    Apr 11, 2011 at 1:38
  • 1
    \$\begingroup\$ Whatever the characteristics of accuracy and reliability of my Pentax Kx, I would be extremely surprised if they were worse than either of the colorimeters I've used since originally posting the question. \$\endgroup\$
    – PeterT
    Mar 5, 2013 at 16:56
10
\$\begingroup\$

Camera Colorimeter is an app on Google Play which uses the camera on your phone as a colorimeter for calibrating other devices.

https://play.google.com/store/apps/details?id=com.auralisoft.colorimeter

[Some experimental results]

I get very consistent capture results using the back camera on a Nexus 6. The following are 10 captured RGB values of the same (rather cool) white screen on a Nexus 7 2013 device, and their standard deviations:

       R             G           B
  0.64074441    0.82363862  0.960373769
  0.637419746   0.823843618 0.960422471
  0.635585636   0.823491139 0.961107378
  0.637959867   0.823703707 0.960575674
  0.637286725   0.823480298 0.960555295
  0.636529085   0.826055671 0.963249426
  0.637193203   0.822952933 0.961079831
  0.635713642   0.824445067 0.962552416
  0.637704785   0.82248129  0.961555964
  0.635952103   0.82470173  0.962591767
  -------------------------------------

Mean 0.63720892 0.823879407 0.961406399

Stdev 0.00142383 0.00094581 0.000989068

I don't have another Nexus 6 to try and a different Nexus 7 device will no doubt return slightly different (but self-consistent) results. We're not interested in what the photos look like or how noisy they may be, but rather whether we can get a good RGB reading averaged over many pixels. Since camera phones have electronic shutters there is not a consistency problem associated with mechanical shutters in SLRs discussed by other posters here.

[Actual results]

Here are some before and after calibration results.

I used the camera on a Nexus 6 to calibrate an Xperia C4 phone, and then measured the Xperia's display before and after calibration using a Datacolor Spyder 4 colorimeter.

The only calibration performed on the Nexus 6's camera was providing it with a reference D65 white point, through the Camera Colorimeter app's "Capture white point" menu option. The reference D65 white point was from another display that has been calibrated using the Spyder 4 colorimeter.

First off is the Xperia C4's uncalibrated gamma curve (yellow):

Uncalibrated gamma

The cyan line is the average gamma. The white curve is the target sRGB gamma. The Xperia C4's uncalibrated gamma is too low and deviates greatly from the sRGB target.

Here is the Xperia C4's gamma curve after calibration, which follows the target sRGB curve much more closely:

gamma after calibration

Next up is a graphical comparison between deltaE (2000) errors for various levels of gray before and after calibration. The average deltaE error dropped from 3.06 to 142 following calibration.

grayscale deltaE before and after calibration

Here are CIE diagrams showing the color gamut (white triangle) and saturation sweeps (small solid circles) before and after calibration. The black triangle is the target sRGB gamut. The small squares are the saturations targets.

Before calibration: CIE diagram (uncalibrated)

After calibration: CIE diagram (calibrated)

Finally, the following plot compares the average deltaE errors before and after calibration for saturation points of individual primary and secondary colors, as well as the overall average deltaE errors across all colors:

Saturation sweeps error comparison

The overall average deltaE error dropped from 3.69 to 2.30.

\$\endgroup\$
3
  • \$\begingroup\$ Why the downvote of this answer? It is a legitmate answer that adds another valid aspect to the issue. \$\endgroup\$
    – user23573
    Jan 4, 2016 at 7:21
  • \$\begingroup\$ Not my down vote, but given the premise of the question, it would be nice to address why this app works (and can be trusted) while in general it is not usually suggested to use cameras in this way. \$\endgroup\$
    – mattdm
    Jan 4, 2016 at 12:18
  • \$\begingroup\$ no wonder you get consistent results, inconsistent results are called noise and they would be clearly visible in your photos. for more thrill, you might check how another nexus 6 sees the same white screen or what values would you read from someone else's nexus 7! \$\endgroup\$
    – szulat
    Jan 14, 2016 at 3:07
9
\$\begingroup\$

There is a very nice step by step tutorial for your idea:

Monitor Color Calibration for free using your DSLR

He adjusts his monitor settings starting with color temperature settings. After that he adjusts the color gain of red, green and blue after taking some more photographs of his monitor showing a color calibration chart.

Everything is done manually. No application was written. The guy who wrote the tutorial uses Mac OS X and a probably expensive RAW image editing software.

You can use Gimp for this, which is free (open) software or RAW-Therapee or UFRaw which are also free (open) software and work with most cameras. They work on all common operating systems.

I am sure the whole process could be improved and refined, but it is a good starting point.

\$\endgroup\$
1
  • \$\begingroup\$ Thanks again for your input. It seems that there is still quite a bit of work to be done by someone to write a tool capable of generating more detailed colour profiles and better automate the process... I've just given you an up vote rather than accepting the answer, I must admit when I looked at it very superficially before I thought there was something cleverer going on. \$\endgroup\$
    – PeterT
    Mar 8, 2013 at 8:21
3
\$\begingroup\$

The basic reason is that the spectral characteristics of a camera are typically not the same as a human observer (A camera with such a characteristic is unusual, and is typically quite expensive and is used for specific tasks such as uniformity testing of displays etc.)

A colorimeter instrument is designed to have a spectral response that matches the standard observer, and measures are taken to ensure a degree of traceable color accuracy. Camera manufacturers want their pictures to look good.

As in all things, you tend to get what you pay for though. Many of the cheap colorimeters aren't terribly good.

\$\endgroup\$
2
\$\begingroup\$

This is actually one very interesting question.

Cons

You can use your camera but the problems are:

  • The camera probably is not calibrated itself. You can choose different white balance settings, I would probably go for the "sun" one.

  • The lens has a color cast by itself. Probably you can shoot without the lens at a very close range. But cameras can have a color cast on the sensor.

  • You still need a reference target. For example, which white are you going to photograph to compare with?

  • The exposition values. Things get more complicated here. How do you know how well is your target exposed?

For a more standardized result, I would choose a very low shutter speed. About 1/2 second.

The link Erik posted is interesting. But it needs to be adapted according to your needs.

  • You need to define the quality or CRI of your illuminant.

Pros

I think a camera can be used in some situations, for example for matching 2 adjacent monitors, using one as a reference.

It can be used to define the white point of a monitor.


But if you do not need a professional solution you can use the normal controllers on your graphics card. I made a basic tutorial here. http://www.otake.com.mx/Apuntes/ColorCalibration/ColorCalibration.phtml

The problem is that you can manipulate these values with just a gamma function.

\$\endgroup\$
1
\$\begingroup\$

I can't give a technical answer but I did this as an experiment - used my DSLR to measure my monitor's color output, after I got frustrated with the skewed and inconsistent results of my dedicated colorimeter. It's a very pedestrian way of calibrating your monitor, but ultimately provided a more accurate solution than the "professional" device (which might have been broken), at least judging by vision and by its consistency when taking more test images.

\$\endgroup\$
1
  • 1
    \$\begingroup\$ I'm tempted to do some open source work to better automate DSLR based calibration at some point... time permitting of course. :) I'm still disenfranchised by the colorimeter thing and reckon good results could be reached. \$\endgroup\$
    – PeterT
    Oct 12, 2012 at 16:42
1
\$\begingroup\$

New here. I've had long "conversations" regarding color management for 16 years with DTPers, professional print shops and photographers. Before digital, I tried color printing, but being a perfectionist, I decided to let the pro labs do the best they could and take the hit on reprints if necessary.

I came screaming and kicking into the digital age. The understanding of light and the changes from red to blue and back during a 24 hour period is essential to understanding color management. The first professionally printed job I did, I went to the printer and asked the layout folks if I could get a copy of their printer profile to embed in the image. They told me it didn't matter what profile I used as long as I converted the image to CMYK. Since the advent of the PDF format, even the need to convert to CMYK is not necessary.

Here's your answer. To use your camera as a colorimeter is not necessary. Yes, as another post stated, the color sensor is biased, mostly to the red end of the spectrum. Others have mentioned visually determining proper color balance. Most people can't tell the difference between Crimson red (Roll Tide) and Cardinal red (Woo Pig Sooie). Most people don't know that light has a color temperature. And that the use of filters can warm or cool the look of the image.

I retrieved some color profiles from the Color Consortium web site years ago and I use those profiles with all of my software and hardware. What I did with the monitor connected to my Windows box was to set the RGB settings to 50% each. Then I made a print and adjusted the monitor brightness to the density of the print (after allowing the ink to dry completely). I then made the necessary "brightness" adjustments, made another print and reset the monitor brightness to the print.

I have been getting acceptable prints from my low-end Epson printers to professional print shops. I have been told by professionals who think that a color profiled monitor, scanner, camera, etc. is a must for professional results. This is true. You must have 5000K lighting (low level), wear black clothing, paint the walls of you workspace 18 percent gray and view the final product under the same 5000K lighting to see the proper color.

Question, does Barnes and Nobel, Borders or any other magazine and book seller employ these conditions? Kinda hard for an outdoor kiosk to manage the lighting so people can view the published colors correctly, don't you know. Sorry for the sarcasm.

The camera is a Reflected Light Recording Instrument. It is not, and was not intended to be, an instrument to measure the color temperature of light for the RLRI operator to make color correction filter decisions. Besides, one can always do a custom white balance under the current lighting conditions to correct the color temperature the camera sees. A colorimeter is a professional instrument to save the working pro time and their clients money by not having to take the time necessary for each location and lighting change to white balance the camera.

Cheers!

\$\endgroup\$
2
  • \$\begingroup\$ "Question, does Barnes and Nobel, Borders or any other magazine and book seller employ these conditions?" I assume you mean book publisher. To some degree, yes, always. And depending upon the publication, yes, to the extreme. \$\endgroup\$ Mar 30, 2016 at 12:58
  • 1
    \$\begingroup\$ No, the lighting conditions where the final product is viewed. \$\endgroup\$
    – kdd
    Mar 31, 2016 at 17:18
0
\$\begingroup\$

There must be some reason why they don't even bother to mention the possibility of using a camera?

  • Inaccuracy?
  • Complexity?
  • Requirement to calibrate the camera first?
  • Inability to properly calibrate the camera for this purpose (cameras are usually calibrated with ColorChecker with reflective colors only)?
  • Inability to eliminate reflections?
  • Too much noise when measuring dark values?

One thing that is notable here is that Datacolor's business model seems to be particularly oriented around not only differentiating their own Spyder products on the basis of the functionality of provided software, but also obstructing the use of the hardware by third party software.

I don't know about DataColor, but X-Rite colorimeters and spectrophotometers can be used with Argyll.

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

Physically, color is a function of a continuous wavelength spectrum. Color calibration needs to work in the colorspace of human eye receptors' spectral responses. If a camera's sensors/filters matched those perfectly, we would not have to bother with white balance to the degree we do.

\$\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.