While understanding the reflectance of scene elements and the camera’s optical response to them is important, to take advantage of the color of infrared, it is important to understand how the camera calculates white balance and what it does to the coloring of converted camera using a red/NIR filter. When a camera records light that one cannot see with the unaided eye, it is electronically translating that light through the three color channels—red, green & blue—into colors on displays that we do see. How it translates, through color balancing, is important, much like translating from one language to another. A normal (unmodified) camera sees red, green and blue about the same way as the naked human eye. An infrared modified camera has responses in the red, green and blue to the near infrared and translates the near infrared colors to your monitor red/gree/blue display. Here is a plot of the near infrared camera color response (sensitivity).
How you perform a custom WB in camera will make for different effects and improve the potential post process results. For example, the following scene--while not the most aesthetically pleasing but illustrative--was taken by custom white balancing off of a frame of just plants with exposure compensation of +1 stop. The plant life in the scene has very little color, while the pond water has a significant reddish tone.
We can examine the histograms of each red, green and blue channel to understand a little more about the scene. Below are shown the Red, Green and Blue channel images with corresponding histograms.
The pond water is brightest in the red channel and darker in the green and blue channels (with pond pixels shown in the red circles of the blue/green histograms). The plant life is about equally bright in all three. This is a demonstration that the three channels have about the same light transmission in the near infrared range where plant life is most reflective.
So if they are equal, then why does plant life generally appear blue-green instead of white? The relative differences in the histograms, as used to compute white balance (color balance) show the reasons. When the red channel is stretched (the max/min pixels pushed to the brightest/darkest values), the pond pixel values are about the same as the plant life values. When the blue & green channels are stretched, the plant life pixel values are pushed high up the brightness scale, while the pond pixel values are pulled to the darkest scale. This contrast difference offers a relative brighter level for green & blue than for red in the plant life pixels, and a relative darker level for pond pixels for the same.
- Saturated colors in sky/water → then WB on plant life.
- Saturated colors in plants → then WB on sky.
A common practice in near infrared color photography processing is the "Channel Swap" method. I will not give a tutorial here, but you can find many by searching online. In summary, the red channel is swapped with the blue so that what is red in the camera data appears blue on the display, and what is blue in the camera data appears red on the display. It actually makes for more pleasing presentation of infrared images.
The above scene is the photo when custom WB of off the plant life, and then processed with channel swap and finally auto-levels (with auto color balance) in Photoshop. It gives very white, colorless plant life, but intensely saturated blue water.
The above scene is the photo when custom WB of off the water, and then processed with channel swap and finally auto-levels (with auto color balance) in Photoshop. It gives dull water color, but saturated plant life. The plant color can be altered using the Hue slider in Photoshop (not described here at this time).
There are other scene elements that can be used for white balance that may give different effects, such as colored cards, building structures (concrete, asphalt) clouds, etc. Experiment with WB and see what brings out different color tones and saturations in your imagery.