|
Datacolor ColorFacts Professional Technical Articles
Understanding Color Measurement
We just explored how a three or four photodetector "controlled condition" color meter can characterize a display for which it has been tuned. We also had a brief discussion of how this tuning can be done through a calibration of the instrument or through a mathematical correction that matches identically the expected output from the device to be measured.
The hypothetical color meter that we are using has a response through the visible spectrum of the following:
This color meter will be fine if it precisely matches the output from the display device, or if it is calibrated to account for the differences. However, it will be accurate only for that specific devices spectral output.
What do you think would happen if we used the color meter with the response pictured immediately above to measure the output of the display pictured here?
If this display were measured with the color meter pictured above, the Red would be significantly incorrect, and the Green would be off as well (look closely at the pictures). Of course, without knowing the spectral output of the display, we might not know if the sensor was working with the display...or not!
In reality, the spectral radiance is hardly ever three defined "spiked" as pictured above. What do you think would happen with the color meter pictured earlier if it measured this display with a very broad spectral output:
There is a lot of color information outside of what the three narrow filtered sensors would detect. Look particularly at the Red region here. It is not terribly "tall", but it is rather "wide". A narrow response color meter would only pick up a small portion of this Red range.
What color do you think the color meter with the response pictured at the top of this page would "see" with a device that output this spectral distribution? What color would your eye see?
The color meter above would indicate that this device has a Cyan (Greenish-Blue) color, as it would not pick up the wide Red range. Even though the sensor says that the color is Cyan, what color is it really?
If you answered "White" to what color this spectral distribution produces, give yourself a big pat on the back! This spectral scan is perfectly dead-on at D65 (x=.313,y=.329), the common reference for white with modern display devices in North America (and elsewhere). Your eye would tell you this, an optical comparator would tell you this, and an expensive spectroradiometer would back it all up.
Isn't that interesting? The color meter that was perfectly fine tuned to the device for which it was created returns a color of "cyan" when measuring this absolutely perfectly white light source. You can imagine what attempting to calibrate with this color meter would do to the image of a display device. Worse, what if you trust this instrument more than your eye? Your calibration could very well make the picture worse. In fact, this is still fairly common.
However, even the broad spectrum above is not the typical output from a normal display, and it is very rare (even difficult) to find very broad sources of light that are this pure. In reality, emissive light sources will have not have three defined "spikes" like higher up on this page, nor a wide broad output across the visible spectrum...
Usually, the spectral output resembles something resembling a profile of the Rocky Mountains, or possibly the Grand Canyon.
If you were to actually see the spectral distribution from a display device, you would likely see large peaks and large valleys, sometimes even changing over time with the age of the display!
Here is a more typical output from a display. This one is from a digital front projector containing a UHP ("Ultra High Pressure") lamp. What color would the color meter at the top of this page see for this projector? What color is it really?
If you guessed "White" on the actual color again, congratulations! This is another display that was calibrated to D65. You can clearly see the Blue and Green spikes that are created from those primaries, but Red is a little bit trickier. What would the color meter show as the color for this display? You are right if you said that it is impossible to even guess on this one without performing the calculations, and since we didn't provide any of the data that goes with these graphs, there is no way to know.
This color meter that is using Red, Green and Blue colored filters could claim that the image appears Cyan as it did in the example above...but it may also return Pink, Yellow, Magenta or Green, or anything else. There is really no way to know, which is perhaps the biggest problem.
So, we have seen that three and four photodetector sensors works sometimes, and can be mathematically massaged to work at other times. What do we need to have a color meter work more of the time?
To have a device that works with all types of displays, we need to precisely match the "standard observer" curves, as mentioned earlier. One way is to build a better tristimulus colorimeter. Another way is to directly measure the radiance within the visible spectrum and calculate what a human would see based on the standard observer curve responses.
Continuous sampling (true integration) of the entire light spectrum would be ideal. Since that is not practical (or economically feasible), discrete sampling of the visible spectrum in finite amounts is a reasonable substitute. Ideally, we wish to sample the radiance of points in the visible spectrum, and from that, calculate what Color the eye would see. Calculating color from spectral radiance is the job of a device called a "Spectroradiometer", and these devices have been providing accurate color information from arbitrary emissive light sources for decades.
Next page.
|
 |
|
| 
| 
| 
| 
