How Does Digital Infrared Work?

How is it that you can put an infrared filter on the front of effectively all digital cameras and get an IR image? In this article we set out to examine how this happens and how to interpret the results.
All digital image sensors are sensitive to more than visible light. They have extended sensitivity in both the UV and IR bands. Because of the IR sensitivity, all modern digital cameras have some sort of IR blocking filter (hot filter) mounted in front of the CCD/CMOS sensor. This will block, to varying degrees depending on how aggressive the camera maker wants to be, some of the IR getting to the sensor. This filter sits in front of the sensor with its integrated Bayer filter of red, green and blue (usually) dyes.

In practice, the hot filter on almost all digital cameras does not completely block IR. This means it is possible to get a shot, just that on cameras with more aggressive hot filters the exposure times can be long. The trend has been to more aggressive IR blocking filters over time. This is why some dedicated digital infrared photographers hang onto older cameras or seek them out secondhand. An alternative with newer cameras is to have the hot filter removed by a camera technician, though it must be replaced with some piece of glass of appropriate thickness so that focusing is not affected. There are companies that will perform this service for you, either replacing the blocking filter with clear glass or with a suitable visible light-blocking filter, like an R72. The advantage of removing the blocking filter are much shorter exposure times that make wedding and portrait photography, for example, possible, as well as handheld landscape work.

Whether you have a camera with its standard IR blocking filter or a modified camera with clear glass inserted, you will need to mount an IR filter on the lens. These filters block all or most of the visible light spectrum so that you are illuminating the camera’s sensor with mostly IR light. A very popular one, and the one I currently use, is the Hoya R72. This filter blocks all the blue and green light but admits some red along with the IR. This is why you can just see through the filter if you hold it up to a very bright scene. Other alternatives, like the Tiffen #87 and the various Wratten filters cut off all the visible light and varying parts of the near infrared. You can expect exposure times to lengthen but the IR effect to become more pronounced as you move to the stronger filters.

Whichever filter you use will preclude using the viewfinder for framing on dSLRs. Compact digitals that have a live preview on the LCD on the back of the camera (and this includes the new Olympus E-330 dSLR) will allow framing using the LCD. Plus cameras with an optical viewfinder will, naturally, allow you to still use that. This can mean that the ‘rangefinder’ digitals, like the Canon G series or the Olympus C family are easier to use than an SLR in some ways.

Two issues that concern many beginner digital infrared photographers are perceived focus and exposure difficulties. In practice, neither is an issue. In all the digital camera cases I have tested,  auto-focus works fine. It is my belief that because the autofocus system uses sensors with effectively the same IR sensitivity as the main imaging sensor, that the cameras automatically compensate for any focus shift. In situations where this is not the case, it seems that the natural depth of field of the camera at shooting aperture handles any miss focus. This may not apply for IR macro photography, though I have not experimented with this to a significant extent, but will for a future article.

Exposure is also not a great issue in practice and something that can be worked out very quickly with a particular camera through a bit of testing. Many digital cameras I have tested will get a correct exposure automatically. Others need some overexposure compensation dialed in. This is the case where the same IR blocking filter does not affect both the exposure-metering sensor and the sensor, and thus there is a mismatch. On such cameras a couple of stops of over-exposure (as indicated) will usually do the trick. On some other cameras, like my Canon 350D, there is such a mismatch between the imaging and exposure sensors that you must use manual exposure mode. This is also rarely a problem. The approach I take is this. Having set the camera on manual and with the camera tripod mounted, I set a moderate aperture (say f5.6) and a one second exposure and take a shot. With the camera set to show the image on the LCD after capture and, ideally, with a histogram superimposed, it is very easy to determine if you have the exposure right. Based on the histogram I adjust the exposure up or down and reshoot until I get a histogram that shows no highlight or shadow clipping and basically a bell curve in the middle of the display. Such an image can be readily improved in Photoshop to make the highlights white(ish) and the shadows black(ish). The brackets are there because you may well not wish to pin them to the maximum and minimum points for aesthetic and printing reasons.

Having locked in a reasonable starting exposure (you may want to adjust this, but we will cover that later), you can happily shoot away. I find that once I have a full sun exposure, say, for a camera, I can simply use that each time in similar conditions or I use it as a starting point in varying conditions. Once you start getting reasonable images you will probably find that the color is red(ish). What is going on?

The color effect that you see in digital IR shots taken with the Hoya R72 and some other filters is a result of the camera still having the red, green, blue (usually) Bayer filter in place. This cannot usually be removed as it is built into most sensors during manufacture. The red color is also caused, if using the R72, by the fact that it allows through some visible light at the red end of the spectrum. While camera manufacturers seem strangely reticent about the topic, sensor manufacturers have to make public some of the information we need. Thanks to Bob McKeever of Kodak Image Sensor Solutions, we have the graph shown below for one of Kodak’s linear sensors. Whilst this is specific to a Kodak sensor, you could expect similar curves from other sensors. You can see that as we move above 750 nm the red filter is letting through all the IR, so the red filter IR sensitivity matches the sensors unfiltered sensitivity. But also both the green and blue filters, which were sitting at effectively zero transmission through the later part of the red spectrum start to admit more light as we move into the IR until they meet the red and allow the sensor to be as sensitive as it would be unfiltered. Now of course if the camera has an IR blocking filter it will be reducing the IR getting through, but this will affect all color channels equally (effectively). The fact that all the color channel filters are effectively transparent to IR above a certain point is the explanation of why many cameras produce a monochrome image when using some of the heavier IR filters.

The faint dotted blue line is the natural sensor sensitivity.

So, what do you do if you are shooting with an R72 or another filter that creates a color image? Well, as you can see from the graph above, the three color channels are capturing somewhat different parts of the IR spectrum. The red channel will be capturing the widest part, the green less and the blue the least, assuming your camera’s sensor follows the above pattern. So the easiest approach is to pick whichever channel gives you the result you want and use that in a convers
ion to monochrome. Or you can just do a monochrome conversion from the color image. Whichever of t
hese approaches you take, you will generally need to use levels or curves to adjust the image. The options to create a color infrared image are explored in the next article.



Lastly, back to exposure. In the screengrab from Adobe’s Lightroom below, you can see that, at the bottom is an intensity histogram, which is what you will see on the back of the camera. It looks pretty reasonable, sitting in the middle of the exposure range. However, when you look at the color histogram at the top, you can see that the red is overexposed whilst the green and blue are in the middle. Depending on your camera you may find that if you intend to use the green or blue channels as the basis for a mono conversion, you may need to overexposure the image as shown on the camera’s histogram to put the green or blue channel nicely in the middle of the graph.


No comments yet.

Leave a Reply