Color shifting with intensity

[Cabman]

We have written a CVI routine which converts the bayer color pattern encoded image from a camera to color. It works fairly well; after we do a white balance the RED, BLUE and GREEN LEDs on our instrument cluster we are testing look fine and consistant.

However, there are some LEDs which are yellow-orange. On the screen they look more greenish-yellow. This would not be so much of a problem, however the color seems to shift with intensity and/or time. As we monitor the LEDs for a while they start to shift from greenish-yellow to yellow. As a result, the matching of color test does not work very well as the values change with time. I'm assuming that the LED light output shifts as it heats up.

We also noticed that the brightest spots on
the green leds look yellow. My guess is that the algorithum does not handle intensity differences well and shifts the color.

Any ideas? Anyone else have this problem?

[Brent_R]

Hello Cabbage,

I'm not 100% certain what's happening here, but I'll make a few guesses for what it's worth. The color shifting over time is probably real. LEDs will change color if the temperature changes. Diode lasers will sometimes have a spec on the datasheet that gives this shift in nm/degC or %/degC. I'm not sure if this is true for LEDs too.

The problem with yellow-orange LEDs looking greenish may be due to the color gains being off. Your Red, Green, and Blue LEDs and the phosphors on monitor are basically monochromatic. The color filter components on the CCD are to a lesser degree. It seems that there might be little coupling beteen the colors when looking at R, G, B LEDs. Even if the color gains are off, red will still look red, green look green,
and blue look blue, however LEDs with a component in multiple bands might not look right.

Most color CCD cameras will provide spectral respose curves in the manual for each color filter and LEDs datasheets should give similar info on their output. It may be helpful to compare the spectra of the LEDs you're measuring with the response of your CCD to see where they fall.

Regards,

Brent R.
Applications Engineer
National Instruments