Equal perceived brightness with different LED colors?

I am creating a front panel which uses discrete LED indicators - red, orange, yellow, green, and blue. Looking on Digi-Key (doesn't everyone?) at the LEDs shows their luminous intensity in millicandelas (mcd), along with their dominant and peak wavelengths.

I would like all of the different colors of LEDs to appear to have the same perceived brightness, and haven't had any success in searching for this. I know the human eye doesn't respond equally to all wavelengths of light, so what do I do next? Would 100mcd LEDs in all colors appear to have the same brightness? Or should I normalize them based on the human eye response?

A separate, but related question: How bright is a 100mcd LED? Is it bright enough to be easily visible when located behind a polycarbonate overlay in muted sunlight? 400mcd? The mcd values are all over the map. I'd like the LEDs to be readily visble, but not painfully bright.

I would prefer to calculate the results rather than just buy a bunch of differently-rated LEDs and empirically determine what is equal brightness. Thanks in advance!

Check out "led human eye response" on the web, plenty of interesting material.
I believe that in the end, having done the background work, you will probably end up having to do some of the work empirically. Photometry is a complex subject, particularly where the human eye is concerned. Instrumentation is available for doing this kind of work, but if it is a simple photometer, you must get calibration charts supplied that show the instrument sensitivity versus wavelength.
Be aware also that there can be a large spread in LED characteristics, if you are going into volume production you may have to get binned parts, for which you will pay more,
good luck,
Richard

Thank you, RIchard.

Sadly, after reviewing a great deal of material (most of which I had seen before), it seems that I will have to get a bunch of LEDs and resistors and empirically determine the best component selection for equal brightness across the colors. I was hoping that there would be a simple way to predict the outcome using specifications and math.

But I would still like to know - is a 100mcd LED bright enough as a starting point? I'd like to narrow my component selection range as much as possible.

Hi Tom,
100mcd may do it, it may not. Unfortunately, it will be an empirical experiment, although you may be able to use one wavelength of LED to do the initial work, and knowing something of the eyes response, calculate approximate output for the other wavelengths.
Are you going to use a diffuse polycarbonate surface, or will the LED be diffuse.
Note the intensity versus temperature graph as well in the data sheet.
One or the other is needed, or the light output will be very directional, of course the beam width plays a part here as well, if you are using a non diffuse LED. The other problem is then background light, like sunlight. You will be aware of that from trying to use any any sort of display outdoors. The dynamic range for clear visibility, even in your "muted" sunlight could be large, is it only going to be on in that condition. If not, then you may have to monitor the brightness of the ambient light, and using PWM , vary the LEDs brightness. That also saves power if the device is not run from the mains.
Without knowing precisely what you are doing, it's difficult to say,
cheers,
Richard