Build your own infrared camera

The unusual tonal values found in infrared photography offer a new perspective on the world. So-called IR cameras are costly, and converting a normal camera to function as an IR camera is expensive. The NoIR camera module for the Raspberry Pi [1] lets you build your own infrared camera.

Theory

Infrared light sits on the electromagnetic spectrum [2] right next to what humans call visible light (Figure 1). Depending on the source, visible light lies between 380 and 700 nanometers (nm). Infrared light occupies the region above and up to 1,400nm. This form of light remains invisible to the human eye, but photo diodes on the sensors of a camera are sensitive to it. As a result, electronics can make infrared light visible to humans.

Figure 1: The infrared region of the light spectrum is not visible to the human eye even though it can be used to create impressive photographic effects. (Image source: Wikipedia)

Every wavelength has a different refraction index. Lenses made from specialty glasses correct these indexes so that images can be captured in focus. This correction process is easier when the image includes only visible light. Moreover, infrared light that has been made visible on an image leads to unaccustomed shifts in color.

Camera manufacturers therefore put a band elimination filter in front of the sensor to prevent infrared light from coming through. The typical camera can only be used in a limited fashion for infrared photography. An additional filter layer, the Bayer filter, is built in alongside the IR band elimination filters. Consisting of red, blue, and green fields, this additional layer turns the light sensors that would otherwise be color blind into color-sensitive sensors. The camera or the RAW converter mixes the color values of the pixels by interpolating the values.

If the pixels of a Bayer filter were porous only with respect to colors, there would be no need for an IR band elimination filter, and there would be no IR photography. However, a Bayer filter lets light pass and overlap in different places along the spectrum depending on the color. As a result, infrared images appear more or less colorful in accordance with the porousness of the color pixels.

A Different Infrared Filter

Although the NoIR module for the Rasp Pi does have a Bayer filter, it does not have one for infrared. The wavelength range that reaches the sensor is therefore significantly larger. The fact that the range is larger means that photographs captured by the sensor turn out with too many red tones (Figure 2). Apparently, the red sensor pixels are the most sensitive in the IR range. However, only the infrared range of the electromagnetic spectrum plays a role in IR photography. Thus, the camera needs an additional band elimination filter capable of filtering out normal light (Figure 3).

Figure 2: The NoIR photograph has far too many red tones because it lacks a band elimination filter.

Figure 3: A 720-nm filter blocks the normal light in this photograph. As a result, the camera captures only the infrared wave range.

Specialty stores carry these filters, often labeling them wrongly as infrared filters. In fact, a filter like this is actually the exact opposite of an infrared filter. The wrongly labeled filter does not block the IR range but instead everything outside of the IR range. Filters come in several versions (e.g., 715, 720, and 830nm); the number assigned to each version equals the maximum wavelength at which the filter will block light. A lower number means that a greater amount of visible light from the upper wavelength range comes through. Theoretically the 7xx filters can be used for fake color infrared photography. The 830nm filter on the other hand only lets information regarding light intensity pass through. Here, the color pixels of the Bayer filter let almost the same amount of light pass through.