How can I evaluate the colour accuracy of my photos?

Question

I have noticed that the colours in my photos are quite dependant on the medium on which I am viewing them — the camera LCD, computer monitor, TV, digital photo frame and print. So, something that looks good on the camera LCD may end up looking quite different on print or even on a monitor, the range can vary quite a bit across the media. Is there any way to better judge the colour accuracy of a photo? Shooting RAW over JPEG should make a difference, I suppose.

Also, along these lines, since the premium lenses are supposed to have better colour reproduction, such a tool\method could also be used to evaluate lenses.

Answer 1

If there's nothing in the picture which provides known, measured reference colors, this is very hard after the fact.

If your image does include reference colors, you can sample them and measure how different they are from the standard. Xrite sells the somewhat-traditional Gretag Macbeth targets, or you can buy more affordable calibration targets produced by Wolf Faust. These targets have a wide range of different colors, because adjusting balance to get reddish-browns more accurate may come at the expense of greenish-blues (for example.)

Even if the image isn't displayed accurately on your monitor, you can use the reference data provided with the color samples to calculate the deviation from ideal for the different colors, as Imaging Resource does in their tests. One would use software like Imatest or DxO Analyzer to do this evaluation. I'm aware of plenty of various other software (including free/open source options) for building device profiles from a reference, but I'm not sure of anything else that gives an analysis of error from ideal.

If you color-calibrate your monitor and use an entirely color-managed workflow, you can be more confidant that the image you see with your eyes represents reality as well.

You may also be interested in evaluating your own ability to judge differences between colors — if you have a high degree of color acuity, you may be more confident in trusting your perceptions (assuming a color-calibrated monitor — or a monitor or print that is the final output), perhaps compared to actually looking at the real scene. There are, of course, serious tests used by eye doctors, but I also recently learned (from here) about a neat online test that you might want to try: the FM 100 Hue Test.

Answer 2

Good looking colors and accurate colors are not exactly the same thing. What most people want is perceptually representative colors, that is the colors as they are remembered, not as they are measured. This is because the mind adjusts colors to account for moderate differences in color-temperature.

The problem of getting such colors is divided into two parts:

• White-Balance: This has to be done on a per-picture (at least per instance of lighting) basis and requires a known reference (as @mattdm mentions). For that you either need an object known to be truly white in the frame or have measured such an object at some point under exactly the same lighting conditions. Because this correction is perceptual, there is no way to have the camera or workflow permanently account for this.
• Color Adjustments: This is where you calibrate the camera or workflow (for RAW) to give accurate colors after taking into account white-balance. Ways to account for this start again from a known reference, usually a color-chart of known color values. This is done once per camera (although as you noticed may have to be done for camera-lens combinations). In case of JPEG, you will adjust image parameters until you get as close as your camera lets you. In the case of RAW, you can generate a profile that converts from the camera's color-space to the exact colors of the color-chart.

Once you have both these steps perfectly, you will know your colors are accurate regardless of the accuracy of any display or print. In order to see a relatively accurate image on particular devices, each device has to be calibrated and the software used to display must rely on an embedded profile (or manually specified, I guess) to show and print your images the way you saw them.

This last point is extremely important as - again - it is a separate thing to have accurate colors in your image files and to see or print accurate colors.

Answer 3

Couple of important fundamentals here:

1.Light is color. Seems simple but when it comes to displays and "color accuracy" that one little factor is key. Chromatic adaption is something our brains do within seconds, so actually perceiving white balance is one of the most difficult things in photographic adjustment. Most have understood that adding a true white surround to their calibrated adjustment process helps this process immensely.

2.) Camera backs can be calibrated and are worth very little in evaluating image quality. Even less if you shoot RAW files since those files are "rendered" using unknown mojo to the cameras display then you get the RAW files and render them using known mojo and have understandably different results.

3.) Colors will never be "Accurate" out of any camera system, calibrated or not. More predictable, or better behaved maybe but never accurate. Camera sensors, as good as they are, do not satisfy the Luther condition (de) by Robert Luther (de) (1868–1945) (also called the "Maxwell-Ives criterion"),because they can not reduce observer metamerism due to the fact that the sensor filters do not match the human eye's cone responses. So give that up and make art. That's what people want from an image. Go for predictability instead. Toss any hope of "Accuracy out the window as impossible.

4.) Calibrate your monitor yes, sure. But if it's in an environment with any natural light understand that your eyes will adapt to that natural light and your monitor calibration is worthless. Natural light changes temp throughout the day causing camera manufacturers to make lots of money on auto white balance functions. So removing natural light and getting a constant 5000k source to view with will greatly improve predictability in your color workflow.