The AP NC series, introduced by Kodak in conjunction with Associated Press in 1994, included an infrared monochrome model, the AP NC2000ir with a resolution of 1012 by 1268 pixels. The DCS 420 and 460 series, introduced in 1995, also included infrared but this time included colour models. The 1999 DSC 6XX series also included an infrared configuration. However, with the basic colour units selling for between ten and twenty thousand dollars, and the infrared versions being special orders on top of that, these were not exactly consumer cameras. The colour infrared cameras, such as the DCS 420, were simply colour digital cameras with the kind of Beyer tri-colour filtering that is still standard, but with the infrared-blocking filter removed.
The aim was to reproduce the false colour abilities of Kodak's EIR infrared Ektachrome and this could be achieved by putting a minus-blue filter in front of the lens. The blue channel then only contained infrared information and the red and green channels contained their colour plus some infrared 'leakage'. By calibrating the system and subtracting appropriate amounts of the blue channel from the other two a result similar to EIR could be achieved.
I'd tried this before, using a borrowed Canon but on a somewhat overcast day. Using my FujuFilm IS-Pro I've shot some more tests, such as the image above, of Eashing Bridge. That's how the raw image looked after each channel had been normalised (ie the darkest and lightest points set).
Using the Photoshop channel mixer, the red output channel is set to the blue input, green output is set to red input and blue output is set to green input. Remember the yellow filter will have removed any real blue from the image. Next some of the blue input channel (which is really the infrared) needs to be subtracted from the green and blue outputs. This is not scientific and is certainly uncalibrated but in this case subtracting 50% blue from the blue output and 65% of blue from the green output gave a suitable colour balance. I also played around with gamma and curves to improve the look of the image: as I said, it's not scientific.
This is the result: