Infrared Photography with a Digital Camera |
Related articles at this site:
Image samples from Olympus cameras: |
It was more than thirty years ago when I last experimented with photography in infrared. Too much hassle: special film handling, black-and-white processing, inability to evaluate results (and adjust settings) until the whole roll was exposed and pictures were printed... | |
Now this has changed. Due to the arrival of digital photography, we can take infrared pictures whenever we please, mixing them with "normal" ones, and see results on the spot, tweaking the settings to our hearts' desires... All depends, of course, on how your camera sensor array reacts to the infrared — and, depending on the filter you are using, to the far red end of the visible spectrum. Note: Any image in a thin frame like this is clickable, linking to larger (XGA) versions. |
Olympus C-5060WZ, Hoya R72 filter |
Camera | E-300 | E-500 | E-410 E-510 | E-30 | E-620 | E-M1 |
Exposure Factor | 7.3 EV (150×) | 10.3 EV (1300×) | 10.0 EV (1000×) | 11.7 EV (3200×) | 11.0 Ev (2000×) | 10.0 EV (1000×) |
Typical Exposure | 1/8 s at F/4 | 1 s at F/4 | 0.7 s at F/4 | 3 s at F/4 | 2 s at F/4 | 0.7 s at F/4 |
From cameras preceding the E-Series: C-5050Z, C-5060WZ, E-10, and E-20 had exposure factor of 11.5 EV (3000×) typically requiring an exposure of 2 seconds at F/4. For other cameras, check for yourself before investing into any filters. You may also find data on selected models at Jen Roesner's site (somewhat dated). A few of the Sony models (Sony DSC-F707/717/818 in this number) allow you to move the anti-IR filter out of the light's way; unfortunately, Sony went to great measures to make this feature unusable in daylight, as it is supposedly making some clothing partially transparent to infrared. (I would also like to turn Sony's attention that the big lens in the '717 can be used to hurt babies and puppies.) Digital SLRs are not my first choice for infrared, at least not off-the-shelf. That's for two main reasons. First, the light metering in digital SLRs is not done by the image sensor itself (like in non-SLR models), but by a separate set of sensors, which may have a different response to infrared. You cannot rely on camera's autoexposure, although you may be able to work out a correction applicable to a given camera/filter combination. Second, many cameras in this class do not offer real-time electronic preview, because the light from the lens reaches the sensor only during the actual exposure. This means that you have to put the camera on a tripod, compose the picture without the IR filter, then put it on and shoot blind. This is less of the problem now than it was when I first wrote this paragraph; after the Olympus E-330, most camera makers offer now models with the Live View. This is why it may be easier to do IR photography using an advanced non-SLR model (a dying breed, indeed), or an electronic-finder camera. But keep on reading. The limited sensitivity of digital cameras to infrared is not the sensor's fault; as I said, it is the anti-IR filter in front of it, intended to improve visible-light images. Therefore a hardcore infrared aficionado may decide to go all the way and use a camera without such a filter. I'm aware of two such cameras on the market, both made by Fujifilm:
Both models are targeted at the forensic/investigative community, and I haven't used either of them, so don't ask me for details. The IS Pro predecessor, called S3 Pro UVIR is no longer offered. Another option is the Sigma SD14. This camera has an anti-IR filter (doubling up as a dust barrier) way ahead of the sensor; it can be easily removed without tools. Recently (early 2008) I've seen the body selling at a mere $600, so the SD14 could be the least expensive way to enter a "full-time" IR photography. Too bad I wasn't able to find any infrared samples from this camera. Last but not least, there are a number of companies which will modify your camera, replacing the anti-IR filter with either a transparent glass (for use in both IR and visible light, with caveats), or with an IR (no "anti") filter. The latter solution allows, in case of an SLR, to use the regular viewing system, as the IR filter is no longer in front of the lens. The conversion is focused on SLRs, mainly (but not only) by Canon and Nikon. Prices range between $350 and $500. At this moment I'm considering a conversion of my Olympus E-500 into an IR-only camera. As soon as I have enough information from the vendor, I will start a new article, documenting my experiences. In the meantime, I will appreciate any Readers who did that sharing theirs with me. The major benefit of a specialized or converted IR camera is a dramatic gain in the effective infrared sensitivity — usually up to 1000× and more; this allows for faster shutter speeds; a matter of not only shooting from hand, but also of higher image quality. More on adapting an SLR camera for infrared can be found in a separate article. The rest of this one is devoted to infrared shooting with a unmodified, visible-light digital camera.
Then you need an infrared filter. You can buy these from any dependable mail-order supplier, like Various filters may differ in the visible light cut-off point (see the table below). The Wratten #89B (available as Hoya R72), with the light transmission falling down to 50% at 720 nanometers, seems to be most popular and gives the greatest chance of success. The darker #87 or #87C may or may not work, depending on the camera, while the almost-IR #70, while allowing for shorter exposure times, does not provide the eerie Woods effect on greens. You also need a way to attach the filter to your lens. This is easy with SLR and digital-finder models, but digital compacts may pose a problem. With very few (like the Olympus C-5060WZ) you can do it directly, as the lens is threaded; with most others you will need a lens adapter tube, like the 41-43 mm CLA-1 attachment for the Olympus C-5050Z (plus a step-up ring).
A tripod is essential. For the #89B (R72) filter you will be getting exposures of 1-2 seconds or more at F/4 and ISO 100. |
Wratten | Schott | B+W | Hoya | Tiffen | 0% | 50% | Remarks |
---|---|---|---|---|---|---|---|
#25 | OG590 | 090 | 25A | 25 | 580 nm | 600 nm | Really a red filter |
#29 | RG630 | 091 | - | 29 | 600 nm | 620 nm | Dark red |
#70 | RG665 | - | - | - | 640 nm | 680 nm | Very dark red |
#89B | RG695 | 092 | R72 | - | 680 nm | 720 nm | Almost "black", but not quite |
#88A | RG715 | - | - | - | 720 nm | 750 nm | I've never seen this one |
#87 | RG780 | - | - | 87 | 740 nm | 795 nm | Cuts off all visible light |
#87C | RG830 | 093 | - | - | 790 nm | 850 nm | Usually called "black" |
#87B | RG850 | - | RM90 | - | 820 nm | 930 nm | Expensive! $250 & up! |
#87A | RG1000 | 094 | RM100 | - | 880 nm | 1050nm | Blocks even some of infrared |
The data above is quoted after W. J. Markerink. For those who would like to have a closer look at IR filter transmittance, here is a graphic representation: | |
The graph shows the transmittance of various filters as a function of the wavelength. It is based on data by Paul Repacholi (Curtin University of Technology, 1992), posted by W. J. Markerink and also by Eric Cheng.
The logarithmic scale is used here, as it fits the problem better: differences, say, between 1% and 10% are by far more meaningful than those between 91% and 100%. It can be clearly seen that #89B (R72) still allows through some of the far red (just below and around 700 nm), #87 starts only around 740 nm. |
Just a reminder: the human eye is sensitive to wavelengths up to 700-720 nm or so. Because the anti-IR filters in digital cameras block most of the infrared, even slight differences in filter transmittance may have strong effect on results (and exposure). Therefore two filters listed as equivalent in the table above may deliver slightly different results on a given camera. Recommendation: I believe the most useful, general-purpose IR filter for digital photography is Hoya R72 (#89B). It blocks visible light well enough (if not entirely) to provide a well-pronounced IR effect, while still allowing for non-exotic exposure times. This filter should work fine with most of mid- to high-end amateur digital cameras (your mileage may vary, so check with someone who tried it on your camera). The small amount of visible (far red) light which this filter lets through does not affect pictures enough to spoil the IR effect, while coloring your images red (or purple), therefore they need to be converted to monochrome in postprocessing. The #87 filter is more expensive and more tricky. Many digital cameras will not be able to "see" through it, while some others may work — again, check with their users. I have tried it with the Olympus E-300: the exposures are much longer than with the R72 (#89B), see here. The filter is now gathering dust at the bottom of a drawer. As mentioned in the SLR section above, most of digital SLRs do not allow you to use the LCD monitor for picture preview and composition. You have to compose in the optical finder without the filter (using a tripod, of course), then put the filter on the lens, and shoot blind. If you have a digital SLR and intend working in infrared, you may be better off getting an inexpensive non-SLR model specifically for infrared. You need a small backup camera anyway. Digital non-SLR cameras (including EVF ones) are more convenient in this aspect, as you can view and compose the image on the monitor as normal; the preview will be red or purple, of course, and quite dark, but usable for these purposes. As mentioned before, the E-330, E-410, and E-510 from Olympus do not suffer from this: their Live View feature allows for electronic image preview with the filter on. All digital cameras I know measure the light through the lens. While this means that the light reaching the metering sensors is already filtered to IR only, there is still dependency on camera type: most non-SLRs use the CCD itself to evaluate the exposure, while current SLRs have dedicated metering sensors for which some of the light used for viewing is diverted. This distinction is quite meaningful. It means that in non-SLR you can usually trust the exposure automation — as long as the metering system is capable of doing its job at very low light levels. It should be: a typical IR exposure with the R72 filter corresponds to exposure value of EV 3, while most cameras can cope with EV zero or close. Even so, however, you will be probably better off applying a negative exposure compensation, usually close to -1 EV, to avoid red channel overload, described below. In SLRs, a separate light sensor does the metering; its sensitivity to the IR may be entirely different than that of the CCD imager (with any IR-blocking filters in front of one or the other accounted for); therefore you may have to apply a significant exposure compensation to get things right. A few test shots should be enough to establish the value appropriate for a particular camera. For example, in bright sunlight I can quite reliably shoot IR with the Olympus E-510 in the autoexposure mode, but I have to apply a +5 EV compensation. Red channel overload. When setting the exposure compensation (SLR or not), you have to aim for a picture which will look like it is underexposed, too dark. This is because practically whole image information goes into just one of the RGB components: red, and you have to keep that component from saturation (i.e., running out of range). If your camera can display a brightness histogram for individual RGB components, make sure that the red one does not hit the upper limit. Otherwise use -1 EV or so of negative exposure compensation, adjusting this correction as you learn your camera/filter combination. | |
Your exposures will be quite long: an IR filter combined with the camera's anti-IR one will let through less than 0.1% of the incoming light. A bright scene, requiring 1/500 s at F/8 in visible light will need about 1 s or longer at F/4 on most cameras. Not only this asks for using a tripod, but, if the air is not quite still, there will be a blur in the foliage, grass, water reflections, etc. This is not necessarily a bad thing, and it may add an extra feel to the image. Here is an example. I like the contrast between the sharp (immobile) planks, and the fluid, fuzzy grass. A matter of taste, of course, but it was quite windy that day. | Olympus E-500, Hoya R72 |
Certain cameras or lenses may exhibit some fogging, or image areas with extra exposure (for example, a bright central spot seen with many Canon lenses). This may be due to light scattered from inner surfaces of the lens, or to some peculiarities of anti-reflective lens coating which was not really designed for infrared. Sometimes the blackness of internal surfaces of the lens tube or mirror chamber may be "not black enough" in infrared. (I experienced the last effect a few years ago with an Olympus E-10, where the extreme 15% or so of many frames was ghosted.) Sometimes it happens to all cameras of a given model, sometimes — just to a particular specimen or a particular lens. Camera makers are not worried about this: very few users ever venture into the IR realm, and this is a mass market after all. There is no way to avoid this problem; once again, check an IR filter on your camera/lens combination before buying. Note to SLR users: regardless of that effect, the image may be fogged, or otherwise affected, by the light entering through the viewfinder in spite of the raised mirror) and reaching the sensor after being scattered around the mirror chamber. To avoid that, close the eyepiece shutter before the exposure, or use the included eyepiece cover (or, at the very least, shield the eyepiece with your hand or hat). Users of non-SLR cameras, obviously, do not have to worry about this. Gisle Hannemyr provides a list of lenses checked to produce a hot spot or fogging (or to be free of these effects) in his Digital Infrared Resource Page. The focal length of your lens (and therefore the proper focus setting) depends on the wavelength. Lens makers try to keep that dependency to a minimum (achromatic lenses), but only within the visible light spectrum. A lens focused in visible light will be somewhat off-focus in infrared. Many film-era lenses had a separate focus marker on the barrel, to be used in infrared photography. With autofocus, like with autoexposure, the outcome depends on the type of camera, although the difference is not as drastic. (Remember, we are talking about non-modified cameras here, where an IR filter is used on the lens.) Non-SLR cameras have an easier job here. Autofocus is performed in the image sensor plane, by contrast detection. This means the circuitry will properly detect when the image is in focus, regardless of the light type. There may be a problem with the amount of light available for the job, but not with its kind; the AF action may be slower and less reliable, but there will be no systematic shift. If your camera is capable of autofocusing in low light down to EV 0 or not much above, you'll be just fine. Sort of. Just in case, I would recommend taking more than one picture, every time forcing the camera to re-focus. This will increase your chances of getting at least one properly focused image. In SLRs the AF is done by dedicated sensors behind a system of mirrors. These sensors are at the same effective distance from the lens as the imager. This should, in principle, work OK, as both the AF sensors and the imager are getting infrared light only. There may be, however, some inaccuracy caused by the fact that both sensors are receiving somewhat different kind of infrared (remember the anti-IR filter on top of the imager!), so both focus planes will be shifted with respect to each other: what the AF sensor will see as in focus, the imager may see somewhat out-of-focus. The difference is not as large as the one between visible and IR light, so we just have to live with this (IR-adapted cameras, without a filter on the lens, face much more of a problem here). Just in case, stepping the lens down for more depth of field, is a reasonable precaution. Some SLRs with the Live View feature offer an additional AF-by-imager mode. In that case, the system works exactly as in a non-SLR camera, just fine — as long as the camera does not do the final focusing, just before opening the shutter, using the "regular" AF sensors. Manual focusing by scale is not really useful here, although with a bit of patience you may find the right setting for a given lens used for landscape shooting. If in doubt, try setting focus a little closer than the actual subject distance: at the equivalent focal length of 50 mm use about 4-5 m instead of infinity — but this actually depends on the particular lens. Depth of field may, to a large extent, help masking the lack of proper focus. In case of problems, try to use wide zoom settings and shoot in aperture priority at F/8 or so; this may help. You can run but you can't hide: at small apertures (large F-numbers) the image resolution is negatively affected by diffraction effects. This effect is more visible (i.e., starts at larger apertures) in infrared. Its magnitude depends on the ratio of the (absolute) aperture size to the average wavelength of light used for the image. That wavelength is about 50% more for near-infrared than for visible light (850 nm or so versus 550 nm; remember that IR wavelength is limited by the anti-IR filter), hence the difference in usable apertures will be about one F-stop, maybe a bit more. This means that if you could (or chose to) use apertures up to F/11 in visible light (with a given lens), in infrared the limit will be F/8 or F/7. Still, it is often easier to live with diffraction than with an out-of-focus image. While smaller imagers suffer from this at lower F-numbers, they also show more depth of field; both effects exactly compensate each other. The response of red, green, and blue photosites in the CCD, and therefore the color image recorded by the camera, is the result of a subtle play between the transmittances of three (!) filter layers involved here: (1) the IR filter mounted on the lens; (2) the anti-IR filter in front of the CCD; (3) the tiny red, green, and blue filters in front of each photosite of the sensor. No wonder that the (false) colors of an image shot in infrared may vary from one camera model to another, and from filter to filter used. While some cameras, especially when used with filters darker than #89B, may come up with color images which some may find pleasing, in most cases the images will have a very strong red or purple-red tint, being recorded mostly in the red component. This is certainly true of all Olympus models I have used, but not only. The effect may also depend on the sensor gain (ISO setting) and color balance, but not by much. Therefore usually you would like to translate your IR pictures into gray scale, i.e., to black-and-white (or sepia) monochrome. Depending on your postprocessing program, you can do it by desaturating the image, or changing its mode to 16-bit monochrome, or applying a duotone filter. After the conversion you will usually want to restore the tonal range of your picture (doing that before does not make much sense, as this usually leads to overloading of the red channel and burning out the highlights). Then, after denoising (see below) and, if needed, some sharpening, you may tint your monochrome picture to sepia or something else, although I usually prefer my IR pictures straight black-and-white. |
Your infrared images do not have to be monochromatic (like black/white or sepia). There is some color information in the image file, and while it has nothing to do with reality, it may be used to generate quite pleasing images. The easiest way to do that is to split your image into individual RGB components, adjust each separately, and then recombine them.
Here is an example of such a treatment, borrowed from my |
Olympus C-5060WZ, Hoya R72 |
Another field for experimentation opens itself if you have two versions of the same image (identically framed): one shot with, and one without an IR filter. Bringing both into an image editor, splitting them into R, G, and B layers, and then recombining the R layer from the IR image with G and B (possibly swapped) from the visible-light one gives effect which is sometimes compared to that of the Ektachrome false-color IR film. Some tonal adjustment of individual layers before recombining them may improve the effect. | ||
The red layer from an IR image combined with green and blue ones from a visible-light picture. |
Frankly speaking, this is too intrusive for my taste, and the effects will be eye-pleasing for very few subjects. The example I'm showing at the left is just plain ugly. You may have more luck (or skills).
In general, I prefer my infrared pictures to be monochrome, sometimes tinted just a bit towards warmer shades. |
Infrared images exhibit more noise than visible-light ones. First of all, the exposures are 1000 times or so longer than in "normal" pictures, and that's when the noise always shows up. Another factor is that the image is created mostly in the red component, i.e., using only the R (red-sensitive) photosites of the image sensor. In the most commonly used Bayer matrix arrangement, out of each four photosites (sometimes wrongly referred to as pixels in this context) one records red, one blue, and two — green component of the image. Only when the raw image is being converted to RGB, the missing values are interpolated, so that a pixel (all three components) is created for each photosite location. This means that your eight-megapixel camera records just two megapixels of information, plain and clear; for the B and G photosites the whole signal is just interpolated from their R neighbors. The photosite size, however, remains still at the 8 MP level, just 25% of what it would be for a 2 MP sensor of the same physical size. Last but not least, the image has to be kept within just a partial luminance range (read: dark), to avoid overloading the red channel, as I already described above. For a more natural look this limited tonal range will have to be, after desaturation, stretched to the full luminance range. This leads to (usually) doubling of the visible noise amplitude. One may argue that the last two effects partially overlap, i.e., I am listing the same thing twice, but still — the noise is there, much more of it than in visible-light images, be it color or monochrome. Cameras with larger sensors show less noise than those with smaller ones (the noise amplitude is roughly inverse-proportional to sensor linear size, at the same pixel count). This is perhaps the most visible advantage of SLRs in infrared applications (assuming that you do not want noise in your IR images, see below). Not because they are SLRs, but just because of larger sensors. Some photographers (especially those who spend most of the time looking at their images in 1:1 pixel scale) are allergic to noise in digital images. Others, however, are so used to the large grain in infrared films, that they consider the noise a part of the "infrared look" and want to emulate the effect in the digital medium. If you dislike noise in your IR images, tough: it will be there. You will have much more of it than in "regular" color pictures. If you find the noise objectionable, try to keep it down by using the lowest ISO setting of your camera and opening the aperture as wide as your lens allows (this will, however, reduce the available depth of field, not always desirable, especially in view of possible focusing problems, already discussed). Remember also that the infrared image should be underexposed only enough to avoid the red channel overload — but not more, as this will result in more noise amplification when the tonal range is restored. If you are not sure, use exposure bracketing until you get the hang of it. Much of the noise can be removed quite well with use of an image-processing program, or, usually more effectively, with a dedicated noise removal application (or plug-in), like, for example, Neat Image. The latter does a very good job, and has a wide range of available adjustments. Still, we are walking a tightrope here. While a moderate amount of noise can be often removed quite nicely, quite often I find that really noisy images look better without that operation, which often leads to excessive loss of texture detail and an unnatural, plasticky look. Be careful and keep the originals in case you change your mind later. Many photographers, to the contrary, like the noise effect, as it is often associated with the "classic", grainy infrared look. Actually, from time to time I'm receiving emails asking how to bring more noise to infrared images! To emulate the IR film grain effect, some writers recommend adding artificial noise in postprocessing. I prefer rather to intensify the natural noise pattern of the CCD. This can be done by setting the CCD sensitivity to the equivalent of ISO 400 (or higher, if your camera allows it). With all cameras I've tried, a one-second exposure at ISO 400 shows more noise than a four-second one at ISO 100, so this will help. As a side effect, it will also reduce the exposure time. To make the noise more visible, you may also underexpose your pictures by one or two EV, and then stretch the tonal range in postprocessing to recover the shadows and highlights; this will also amplify the noise considerably. This technique takes less time to apply than to describe, see an example below. Just remember that the tonal smoothness of the result may be affected, therefore it may be better to convert your image to 48 bits (16 bit per color), do the processing, and only at the end convert it back to 24 bits. The noise amplitude will remain unaffected by the switch to 48 bits, but the tonality will be smoother. |
Olympus E-20 at ISO 320, Hoya R72, underexposed and re-equalized after desaturation |
A 1:1 sample from the image at the left |
Olympus C-3000Z; Wratten #70 |
A deep-red filter may still provide very nice monochrome images, as long as you do not have to have the Woods effect.
I'm using it from time to time, and the extra advantage is that with the exposure multiplier you can get away shooting from hand. Some examples are shown in my C-3000 almost-IR sample page. |
Olympus C-5060WZ, Hoya R72 filter |
Infrared is not a common medium for people shots, but I have seen some very good IR portraits. Outdoors portraits or nudes with the surrealist background may have a strong impact. The skin becomes more white, with most imperfections (and texture) gone, which may be related to the partial transparency of its outer layers to the longer wavelengths, and the porcelain-like tonality can be quite interesting. Check out, for example one of Eric Cheng's IR galleries (shot with a modified camera, with the IR-blocking filter removed). | |
Back in 2000, when this article was originally posted, there was a scarcity of any IR information on the Web (Warren, McCreary, and Cheng being notable exceptions). In following years more photographer/writers started discovering this area, so there is more information available now. Here are articles which attracted my attention, in the chronological order.
|
Check also The Beauty of Infrared page in my Gallery section. |
Home: wrotniak.net | Search this site | Change font size |
Posted 2000/07/17; last updated 2013/12/14 | Copyright © 2000-2013 by J. Andrzej Wrotniak |