|
Digital Photography For What It's Worth |
![Fallen leaves under IR lighting; tripod, Canon D30, Hoya R72, f/5.6, 8 sec, ISO 100. Click for 640x428. Courtesy and copyright Chris Mieker; all rights reserved. [Canon D30]](http://dpfwiw.com/images/fallen_leaves.jpg) |
Conventional visible light photography is challenging enough. Why bother with infrared? Because it opens up an otherwise unseen corner of the world — one of serene beauty and never-ending surprise. Digital cameras make this peek around the red end of the visible spectrum easier than ever before.
On this page...
![Formal gardens at Washington Park, Denver, CO; R72 IR filter with monopod support. Click for 800x600 version. [C-2020Z]](http://dpfwiw.com/images/r72_gardens.jpg) |
Denver's Washington Park and the Rockies beyond |
Curiosity about the world beyond natural perception motivated some of our greatest inventions and scientific advances. Since the 17th century days of Galileo and Leeuvenhoek, telescopes and microscopes working with visible light have extended the reach of human vision to ever larger and smaller scales — today by many, many orders of magnitude. To say that these instruments have revolutionized all of science and much of Western philosophy and even religion in the process would not overstate the case.
In the last 2 centuries, visual observation escaped not only the human scale but also the visible spectrum — that narrow band of electromagnetic (EM) radiation where the solar power spectrum and the sensitivity of the human eye both peak. (Certainly no coincidence there.) Cameras and films sensitive to infrared and ultraviolet light gave us our first glimpses of a world awash in invisible light. Imaging devices based on more exotic forms of light (X-rays, radio waves, etc.) soon followed and continue to proliferate.
Today, as ever more sophisticated observing devices open up new segments of the EM spectrum to our view and analysis, astronomers and cosmologists find it necessary to revise their understandings of the cosmos — and even of our own solar system — on an almost continuous basis. And along with these views of the world beyond the senses have come many scenes of unimaginable beauty. Any image at Hubble Space Telescope or in the W. M. Keck Observatory gallery will attest to that.
 |
Saturn's moon Titan at 0.8-5.1 microns (near to far IR) as captured by Cassin on 10/26/2004 from an altitude of ~450,000 kilometers (280,000 miles). |
Out-of-spectrum experiences have generally been beyond the reach of the average photographer, but today's silicon-based consumer-grade digital cameras make it easy to explore the strange and serene corner of the invisible world found just beyond visible red in the near infrared (NIR) band of the EM spectrum at 700-1200 nm (0.7-1.2µ) wavelengths. Throughout this article, the terms infrared, IR, near IR and NIR will refer to the 700-1200 nm band of interest to digital photographers unless otherwise noted.
 |
Twin Keck telescopes atop Mauna Kea |
To this day, the NIR remains one of the most useful extra-visible bands in the EM spectrum. Aerial photographers have long relied on NIR imagery to capture the landscape with the greatest possible clarity over a wide range of atmospheric conditions — including some quite unsuitable for visible light photography. For much the same reasons, the world-class Keck telescopes atop Mauna Kea (right) spend much of their precious observing time with sophisticated digital NIR detectors mounted. Abundant interstellar NIR radiation conveniently passes through dust, gas and our own atmosphere to allow glimpses into otherwise hopelessly obscured regions like the Milky Way's galactic center.
At left is one of humankind's first-ever looks at the surface of Titan, one of Jupiter's four large Gallilean moons. The 0.8-5.1 micron infrared wavelengths were chosen specifically for the Cassini flyby in order to cut through the haze that completely obscured the surface to the Galileo flyby a decade earlier. Titan is in many ways a frozen version of Earth.
NIR reflectance patterns |
|
![Lower Lake, Washington Park, Denver, CO; visible light, 1.7x teleconverter. Click for 800x600 version. [C-2020Z]](http://dpfwiw.com/images/park_vis.jpg) |
![Lower Lake, Washington Park, Denver, CO; R72 IR filter and 1.7x teleconverter with monopod support. Click for 800x600 version. [C-2020Z]](http://dpfwiw.com/images/park_r72.jpg) |
Visible Light |
Near Infrared |
The false color schemes seen in digital IR images like the park scene at right and the leaf still-life at top are another matter entirely. The colors are nothing more than artifacts deeply rooted in camera hardware and firmware, with no direct connection to the objects imaged. Colors aren't even defined in the NIR, of course, but the false colors can add their own mystique to digital IR photographs, and some digital IR photographers like Chris Miekus work hard to manipulate them to their own ends.
In all fairness, NIR images aren't for everyone. As accomplished film IR photographer Josh Putnam once put it on RPD, "... people either love [IR photography] or just don't get it, but the ones who love it really love it." That seems to be more true of photographers than of viewers, however. IR photographers commonly find that their IR images generate more interest than their visible light images. Many people appreciate IR's fresh view of things, but it's not just a matter of novelty. IR images have a rich beauty all their own.
Unlike ordinary films, silicon-based CCDs and CMOS sensors turn out to be quite sensitive to the near infrared (NIR) in the 700-1200 nm (0.7-1.2µ) range — so much so, in fact, that some of the incoming NIR has to be filtered out in order to reduce IR contamination artifacts to acceptable levels in the visible light images most buyers aim to take. The usual solution is to fit digital camera sensors with special internal IR cut filters (IICFs). These sensor-mounted filters vary in their IR transmission spectra, but most consumer-grade digital cameras let enough NIR through to allow some IR photography. Despite a clear trend toward ever-lower IR sensitivities in higher-end cameras, that's still true in 2004, but it gets harder with every passing year. If you get hooked on digital IR, you may end up searching high and low for an Oly C-2020Z or Nikon Coolpix 900, or for the more recent 5MP Minolta Dimage 7. These discontinued cameras are all still quite competent by any standard, but the high prices they command largely reflect their extraordinary IR capabilities. No, my C-2020Z isn't for sale.
Special external filters passing NIR while blocking most or preferably all visible light make infrared photography possible. Film-based IR photographers have been using such filters for decades, but daunting technical and financial challenges continue to keep IR film photography well out of the photographic mainstream.
Luckily,
digital cameras have changed all that.
Armed with even an inexpensive IR pass filter, a sufficiently IR-sensitive digital camera
makes IR photography
Eminently affordable — no need for expensive IR film and developing,
Enjoyably impromptu — no need to switch back and forth between regular and IR-sensitive films,
Easily learnable — via immediate feedback in the field,
Immensely fun — for the many surprises the NIR world holds (and for all the reasons above), and
With quality IR pass filters like the Hoya R72 going for as little as $24, there's hardly a good reason not to try IR if your camera's up to it.
In late 1999, dpFWIW contributor Jay Scott shared with me his first forays into the IR realm with an Oly C-2020Z:
I got a Hoya R72 IR filter, and I love it. Taking infrared pictures could not be easier. Put the filter on the lens, optionally set the camera to black-and-white, and go. The camera has enough sensitivity to IR that daylight IR photos can be taken handheld, and I was even able to get indoor IR photos under incandescent light. White plants, dark sky—I love it!
A few months later, Jay was still at it, more enthusiastic than ever.
IR photography with a digital camera is almost like having an IR eye. You point the camera and look at the display, and you see what it sees. It's so much fun to peer at this surreal invisible world that I wonder why everyone doesn't!
Most infrared photos are made either outdoors or in a studio. But incandescent lights are IR-bright, and I found it easy to make IR photos in ordinary indoor lighting. You have to be prepared for exposure times up to one second, depending on how bright the room is, but it's not a problem if you have camera support. Because incandescent lights are redder than daylight, the "infrared effect" is stronger indoors, at least with my Hoya R72; you are effectively photographing using longer wavelengths than you would outdoors. Household objects can surprise you with their weird appearance in IR; my blue dishwashing soap turned out to be IR-transparent.
Plants are white in IR. Flowers are bright, but seed-heads are often dark. The sky is dark, but clouds are bright. Skin looks strangely smooth, which could be an advantage for some portraits. Any object which is hot enough to glow red is more than hot enough to glow near-infrared. I've photographed gas flames and hot coals.
The camera's flash is bright enough for IR macro photography with the R72. A decent external flash should be bright enough to take pictures at portrait ranges. Someday I hope to get an IR flash head so I can take IR photos at night without being noticed.
Let me emphasize here that digital IR photography typically relies on reflected NIR from sources like the sun and incandescent lamps. Digital camera sensors based on silicon are not sensitive to the far (thermal) IR wavelengths (typically 3.0µ and longer) emitted by objects at room to body temperatures. Heat leaks from houses aren't visible in the NIR, and people, animals and other objects at room to body temperatures don't glow in the NIR any more than they do in visible light. To photograph them in the dark, you have to provide proper NIR illumination using a suitably equipped camera like the Sony DSC-F7x7 or an external NIR-only flash with no filter.
This article doesn't have much to offer on NIR-illuminated night photography, but many other web sites only a Google search away address this rich and useful field.
Digital and film IR photographs have a look many describe as surreal. Clear, serene, bold and tonal are additional words that come to mind, at least for landscapes. Physical and firmware-related factors contributing to the IR look are discussed below. Aerial and reconnaissance photographers have long valued the often stunning clarity characteristic of IR photographs, and it tops my list of IR virtues as well.
Clarity comparison |
|
![Vollmer Peak on a hazy afternoon, Orinda, CA; no filter. Click to see 800x600 version. [C-2020Z]](http://dpfwiw.com/images/hazy.jpg) |
![Vollmer Peak on a hazy afternoon, Orinda, CA; Hoya R72 IR filter. Click to see 800x600 version. [C-2020Z]](http://dpfwiw.com/images/hazy-ir.jpg) |
Visible light |
Near infrared |
IR images owe their great clarity to the atmosphere's exceptional transparency in the NIR. Scattering by air molecules is much less efficient at NIR than at most visible wavelengths. As a result, NIR photons take on average a much straighter path from object to CCD.
In visible light (left), scattering severely limits detail on the more distant portions of the far hillside in this hazy afternoon scene. Removing visible light with a Hoya R72 IR filter takes out much of the detail-scrambling scatter. An impressive amount of detail shines through the haze in the IR image on the right, despite the odd false-color scheme.
Usually monochrome or nearly so, IR images also partake of the deeply tonal beauty typical of black-and-white photographs. In combination, these visual charms make for some truly stunning IR images.
Before rushing out to buy an IR filter, test your camera to make sure it can do its part.
On this page
 |
Warning! Some digital cameras are better suited to IR work than others, and a few are downright hopeless. |
Ever since 3 MP CCDs hit the consumer scene in late 1999, digital cameras have varied widely in their IR performance, but the overall trend has been toward lower and lower IR sensitivities ever since. As of 4Q2004, quality handheld IR photos are unlikely with most new cameras — even through an IR filter with minimum light loss like the popular R72. But if you come prepared for long exposures, in-camera or post-processing noise reduction and rock-solid camera support (read "tripod"), even relatively insensitive cameras can produce satisfying IR photos.
If you're looking for a digital camera specifically for IR work, you'll have to buy used. The 2002-vintatage Minolta Dimage 7 or 1999-vintage cameras like the Oly C-2020Z, the Nikon CoolPix 950 and the Nikon D1 all good bets.
With few exceptions, the very best IR cameras ever — most notably, the Oly C-2020Z, the Nikon CoolPix 950 and D1, and the Canon Pro70 — came out ca. 1999, and most were built around Sony's then-popular 2.11 MP CCD. The 5MP Minolta Dimage 7 (D7) was a surprise late-comer to the club. All had some combination of the following features:
CCD or CMOS sensors with large sensels — through either large sensor diameters (the D1 and D7) or fewer sensels crammed onto average-sized chips (all the others)
Less restrictive internal IR cut filters (IIRCF) — presumably through less manufacturer paranoia about IR contamination in visible light photos (the C-2020Z)
Fast lenses (especially the D1 and C-2020Z)
Flexible exposure, including manual
Effective in-camera noise reduction at high ISO (the D1)
Firmware that happened to render IR images in pleasing false color schemes (the C-2020Z and Nikon 950)
Oly's 2nd generation digital rangefinder, the C-2020Z, achieved its legendary IR performance through a combination of large 0.0039 mm sensels, a permissive IIRCF and a fast f/2.0 lens. With an R72 IR filter, it suffered only 5-6 stops of light loss and could easily produce crisp handheld IR shots on sunny days at ISO 100. That's become the de facto gold standard for IR sensitivity among consumer-grade digital cameras, and that's the standard I'll use when comparing all the other cameras mentioned here. And for all that IR sensitivity, the C-2020Z still took great visible light photos with no noticeable IR contamination in the vast majority of situations. It also rendered IR photos in a false-color scheme that most users could live with and many came to like.
![Handheld C-5050Z/R72 image in the camera's native false color scheme [C-5050Z]](http://dpfwiw.com/images/041029-1714-56e.jpg) |
Handheld C-5050Z/R72 in the camera's native false color scheme |
No digital camera sold new or even factory-refurbished in 2003-2004 can match the C-2020Z's IR performance. A typical case in point is the C-2020Z's 5th generation descendent, the Oly C-5050Z, which I purchased as a factory refurb in 4Q2003. The C-5050Z is a very highly regarded camera in its own right, but it's IR performance is so-so at best. Smaller 0.0028 mm sensels and a much more restrictive IIRCF render the C-5050Z a whopping 5-6 stops less IR-sensitive than the C-2020Z. With an even faster f/1.8 lens and reduced noise at maximum ISO (400), the C-5050Z manages to stay in the IR game but only with the R72. The deeper 87-series filters are out of the question. At right is a handheld R72 taken at maximum ISO and aperture with shutter speed fixed at 1/20 sec (which I can usually keep still).
In theory, the C-5050Z's less IR-friendly IIRCF reduces IR contamination in visible light work, but in practice, I can't say I see the benefit, even against the very IR-sensitive C-2020Z. The table below details the differences between these two cameras and how they contribute to their respective IR sensitivities.
IR Performance Comparison: Oly C-2020Z vs. C-5050Z |
|||
| Property | C-2020Z | C-5050Z | Comments |
| Sensel size (mm) | 0.0039 | 0.0028 | These sensel diameters translate into 94% more sensel area in the C-2020Z. |
| Widest aperture | f/2.0 | f/1.8 | The fast C-5050Z lens gives it a 0.3 stop edge, but sensel area is more important. |
| Noise at ISO 400 | Moderate to severe | Mild to moderate | Increased quantum noise due to smaller sensels partially offsets the reduced noise of the C-5050Z's electronics. |
| Remote test |  |
 |
A remote test result as dim as
the C-5050Z's doesn't necessarily put the camera out of the IR game, but
it does mean that IR work will require solid camera support and long
exposure times far beyond the handheld range.
Technical Note: For a straight-across comparison, both cameras were fixed at f/2.8 @ 1/20 sec, ISO 100 in manual exposure mode at full zoom (105 mm EFL). The full-sized test images were cropped without further post-processing. |
| Handheld sample :
Visible light |
 |
 |
In visible light,
the two cameras are equally sensitive, as one can see from their nearly
equal EVs, which have been normalized to
ISO 100.
Technical Note: For the handheld samples in this row and the next, both cameras were set for maximum zoom (105 mm EFL), auto-exposure and auto-WB with in-camera sharpening disabled. Post-processing of the full-size sample images consisted only of resampling to 800x600 and unsharp masking (USM) applied the brightness channel in Lab mode with identical USM parameters. All images were saved with identical JPEG compression settings. |
| f/6.3 @ 1/500 sec, ISO 100 | f/4.5 @ 1/800 sec, ISO 64 | ||
| EV 14.3 | EV 14.6 | ||
| Handheld
sample:
IR with R72 filter |
 |
 |
The C-2020Z's faster
shutter speed eliminates camera shake
and partially stops the motion of the passing cyclist. It is just this
kind of IR performance that makes the C-2020Z a perennial favorite among
IR enthusiasts.
Note also the strikingly different false-color schemes these cameras assign to identical R72 scenes. Differing auto-WB algorithms probably account for much of the disparity. One would normally apply an IR-specific WB setting to improve the C-5050Z's garish rendering. Luckily, the C-5050Z has that capability; the C-2020Z doesn't, but many find its false color scheme rather pleasant. |
| f/2.8 @ 1/40 sec, ISO 100 | f/2.6 @ 1/5 sec, ISO 400 | ||
| EV 8.3 | EV 3.1 | ||
| Sensitivity loss in the NIR band | -6.0 stops | -11.5 stops | Bottom line: The C-2020Z is 5.5 stops more IR-sensitive than the C-5050Z under these test conditions. |
Silicon-based (CCD and CMOS) image sensors are equally sensitive to visible and NIR wavelengths out to about 1200 nm. To fashion a peak sensitivity in the visible band and to minimize IR contamination of visible light images, most if not all digital camera manufacturers cover their silicon sensors with an internal IR cut filter (IIRCF). Little is known about them outside the camera manufacturers and their suppliers, but some IIRCFs are clearly more restrictive than others. It's often stated, quite incorrectly, that professional cameras usually have very restrictive IIRCFs and consumer models usually don't, but the empirical Nikon D1 vs. 990 IR sensitivities noted below tell a very different story. The pattern — if there is one — remains unclear.
Some hardcore digital IR enthusiasts have gone so far as to disassemble their cameras and remove the IIRCF, a move guaranteed to increase IR sensitivity and void the warranty. James Wooten's Removing the IR Blocking Filter in the Nikon CoolPix 990 and 995 nicely illustrates the results and the surgery, which involves replacing the IIRCF with a plain piece of glass to preserve the camera's auto-focus ability. The IR galleries at Don Ellis' www.kleptography.com include many shots captured with a similarly modified Canon G1. Few have Don's eye for IR images.
With built-in NIR illuminators and a "Nightshot" mode that removes the IIRCF from the lightpath to the CCD, Sony DSC-F7x7 digital still cameras excel at IR-enhanced low-light work. They would seem to be naturals for daylight IR work as well, but Sony felt a need to force auto-exposure metering and to restrict Nightshot exposures (f/2 at 1/60 sec or longer) to keep voyeurs from subverting Nightshot to see through clothing during the day. (Some fabrics are apparently transparent enough in the NIR to reveal what's underneath in bright sunlight.)
These firmware restrictions pose challenges for legitimate daylight IR work with F7x7 cameras, to be sure, but with ISO locked at 100, a deep IR filter (e.g., the Wratten 87c or the Hoya RM90 or RM100) and one or more ND filters to cut daylight NIR input, the F7x7s can rise to the occasion, as Paul Cordes recently detailed on RPD:
I'm using a B+W 58 093IR, equivalent to an 87c. The filter is absolutely black and admits no light to visual inspection. The pictures are great.... You'll also need some Neutral Density filters, as the 707 is limited to f2 and 1/60th sec as the fastest shutter speed (Nightshot Mode). I'd suggest 0.3, 0.6 and 0.9 ND filters to give you flexibility for exposure control. If you don't mind a 3-filter stack, you can skip the 0.9 ND. Don't forget to block the IR emitters with something IR-opaque, or you'll see reflections from the back side of the filters.
Stacking filters of course results in vignetting. If you think that you will be stacking, a step-up ring and IR and ND filters of, say, 67mm might be a good starting choice. Wish I had thought of that before I bought mine!
As usual, testing is better than assuming. Here's a very
crude
test of IR sensitivity for digital cameras:
Put your camera in program mode at ISO 100.
Point a TV, camera or other IR remote into the lens from no more than 12" away.
Press any button on the remote.
Remote Test Comparison |
|
|
|
C-2020Z |
C-5050Z |
If the beam is as dim as the one captured at left with my Oly C-5050Z, you can still get IR images with an R72, but handholding will be iffy at best. You'll generally need solid camera support and long exposure times.
If your camera passed the remote test, it's almost certainly good to go with a Hoya R72 filter or Wratten 89b equivalent. But can it also handle the more expensive black IR filters like the Wratten 87 and 87c? The C-5050Z certainly can't. Compatibility with 87-series filters is hard to predict reliably with the remote test alone, but you can at least get an idea if you happen to have access to another camera of known 87 compatibility.
If you can't compare with a known reference, consider starting out with a Hoya R72 or equivalent Wratten 89b filter. (FWIW, the R72 is still my favorite IR filter overall because it can be handheld most of the time on my C-2020Z.) If your camera sustains less than 7-8 stops of light loss in bright sunlit scenes with an R72/89b, it has a very good chance with an 88a and at least a fighting chance with an 87 filter. If it loses more than 7-8 stops, it's not likely to fare well with a black 87, but it still has a chance with the shallower 88a.
Included in this category are cameras that come within 3 stops of the Oly C-2020Z when used with an R72 filter. As noted above, the most IR-sensitive higher-end consumer-grade digital cameras were built around the large-sensel original Sony 2.11MP CCD ca. 1999. The Oly C-2020Z, Oly C-2000Z and the Nikon CoolPix 950 all used this CCD, and all do well with deeper IR filters like the Wratten 87 series. The red-filtered sensels in this now-obsolete CCD approached the 700 nm visible-IR boundary with a whopping ~70% residual sensitivity and a relatively shallow drop-off into the NIR. That left a generous window for NIR recording, particularly for shortwave (700-770 nm) NIR. The updated 2.11MP CCD found in the Oly C-2040Z is much less IR-sensitive, as noted below.
I've also seen excellent IR images taken with the 2MP Canon Pro70 using an R72-equivalent Kodak Wratten 89b gel, but I don't know what CCD it used. By all accounts, the IR sensitivity of the large-sensel 2.5MP Nikon D1 also seems to fall in this category, but I know little about D1 internals. So much for the theory that professional digital cameras have more restrictive IIRCFs.
The 5MP Minolta Dimage 7 (D7) bucks the generally valid inverse relationship between pixel count and IR sensitivity, thanks in part to the large sensels on its 2/3" type CCD. Based on exposure data provided with a number of online D7 IR images taken with an R72 or equivalent filter, the D7 appears to run only 3 stops or so behind the C-2020Z in IR sensitivity. Soon after the D7 came out, Minolta released Dimage 7i ahd 7hi successor models, both with much more restrictive IIRCFs and correspondingly much weaker IR performance. If you're interested in high-resolution IR work using a camera with a more current feature set, consider the D7 but stay away from the D7i and D7hi.
IR performance can be hard to predict. When physicist David Brown conducted an informal IR comparison pitting an Oly C-2040Z, C-4040Z and E-10 against each other using an R72 filter and auto-ISO, the C-4040Z produced the best IR images by a wide margin, not by virtue of greater IR-sensitivity, but via lens speed and superior performance at high ISO. At an automatic ISO of 337, his handheld C-4040Z turned in shorter exposures and crisper R72 images. He also found the C-2040Z to considerably less IR-capable than his old C-2000Z. These findings jibe with Oly's claim that the C-2040Z's 2.11MP CCD had been redesigned since the C-2020Z and C-2000Z (and apparently since the C-2100UZ), but the difference could have been nothing more than a more restrictive internal IR cut filter.
Also much to my surprise, Roland Karlsson reported on RPD that his Heliopan RG 780 (Wratten 87 equivalent) and Heliopan RG 850 (Wratten 87b equivalent) IR pass filters both work well with his 3.34MP Sony S70. Since the RG 850 has an even deeper NIR window than the Wratten 87c, the entire 87 series should work with the S70. The S70's substantially greater IR sensitivity relative to most other cameras (the Oly C-30x0Z, Nikon 990, Canon G1, etc.) with the Sony 3.34MP CCD points to a less restrictive IIRCF.
In Nightshot mode, Sony's DSC-7x7 cameras remove their IIRCFs from the lightpath to the CCD in order to capture low-light scenes illuminated with NIR. If it weren't for firmware restrictions designed to thwart voyeurs, their 7x7 cameras would be the most IR-sensitive around by a wide margin.
Included in this category are cameras
more than 3 stops less sensitive than the Oly C-2020Z when
used with an R72 filter.
For better or for worse, higher-end consumer-grade digital cameras have been getting progressively less IR-sensitive since 3MP CCDs hit the market in 1999. If you lay awake at night worrying about IR contamination in your visible light images, that's a feature, but if you hope to get in some digital IR photography with one of these cameras, it's a bug.
Many of the early 3MP cameras introduced in 1999-2000 — the Nikon CoolPix 990 and 995; Oly C3030-Z, c-3020Z and C-3000Z; Canon PowerShot G1 and IS Pro90 — were based variations of the Sony 3.34MP CCD. All ran at least 3-4 stops less IR-sensitive than the Oly C-2020Z, as you can see from R72 exposure data gleaned from Don Ellis' IR gallery (Canon G1) and Todd Walker's R72 samples(Canon Pro90). IR buffs who tried them with deeper IR filters like the Wratten 87 series and the Hoya RM1000 eventually gave up, but those willing to put up with tripods and long exposures had considerable success with the more forgiving R72/89b filters. Don certainly didn't let the challenges get in his way.
The Minolta Dimage 7 aside, subsequent cameras with 4MP and higher resolutions have on average been even less IR-sensitive than their 3MP forebears. My 5MP Oly C-5050Z, a typical example, runs 5-6 stops behind the C-2020Z through an R72. To understand why and what that means for use in IR work, see the detailed IR-oriented comparison of the C-5050Z and C-2020Z above.
![Typical Canon D-30 false color scheme for the R72 on auto-WB. Click to see 640x320 version. Courtesy and copyright Chris Miekus, all rights reserved. [Canon D30]](ir_files/d30-r72_colors0.jpg) |
D30/R72 false color scheme |
Chris Miekus, an RPD regular with an enviable knack for non-landscape IR work, took the featured photo at the top of this article with a Hoya R72 mounted on a Canon D30 digital SLR with a Canon EF 28-135 mm, f/3.5-5.6 IS lens at f/5.6, 8 sec and ISO 100 (EV = 2.0). A common "heat lamp" provided the delicate lighting for this indoor subject. Chris reports that with auto white balance, the D30 produces a subdued magenta R72 false color scheme similar to that of the Canon G1, as shown here. For the photo at top, however, he applied a custom white balance to the RAW D30 recording after the fact.
The D30 appears to run 4-5 stops less IR-sensitive than the Oly C-2020Z. According to Chris, the D30 is more IR-sensitive than its immediate successor, the D60, and about 3 stops more sensitive than the more recent D100 offering.
Chris cautions that lens choice is critical in IR work with Canon DSLRs. The popular Canon EOS 50 mm f/1.4, EF 16-35mm f/2.8 L and EF 28-70 mm f/2.8 L lenses all have anti-IR coatings that create bright central artifacts. As of 3Q2003, I know of no other IR-incompatible Canon lenses.
Many newer digital cameras still retain enough IR sensitivity for patient tripod-based IR work with a dark red R72 filter or equivalent Wratten 89B, as Don Ellis' IR gallery attests. Some even appear able to venture a bit deeper into the NIR. Danny Gossens reported on RPD that his G1 works well with a Heliopan RG715 filter — a Wratten 88A equivalent with a ~750 nm 50% cut-off falling between the Hoya R72 (Wratten 89B) and the Wratten 87 series, according to the spectral data published here. Note that Danny's posted G1 IR samples all required tripod support and shutter speeds of 1/3 sec or longer, but his results are more than acceptable.
Specially-designed filters that block everything but IR make IR photography possible. Here we discuss the main types and some of the properties that distinguish them.
On this page...
![IR view of Mt. Tamalpais, looking NE from Vollmer Peak; monopod, Hoya R72. Click for 800x600. [C-2020Z]](http://dpfwiw.com/images/tam_ir.jpg) |
Mount Tamalpais from the Berkeley-Oakland Ridge, Northern California. Thanks to Nemus Photography of Lake Tahoe for the Location tip. |
Let's pause to clarify some potentially confusing filter terminology. The filters used for IR photography are commonly referred to as "IR filters". To the extent that the much more common filters blocking UV light are properly called "UV filters", that may seem something of a misnomer, but like it or not, this usage is firmly entrenched in the IR band.
|
In this article, I'll refer to filters that block IR and pass visible light as "IR cut filters", as opposed to the "IR filters" or "IR pass filters" that enable IR photography by doing just the opposite. |
Among the many IR filters available, two are particularly noteworthy:
In glass in the 49 mm size, these filters went for $25 and $83, respectively, at The Filter Connection as of April, 2000.
 |
Data from Clive Warren's Infrared Photography FAQ and the Heliopan No. 8125 Digital filter page. |
The graphs at right show transmission spectra for several popular filters — the Wratten 89b (R72), 87 and 87c (B+W 093) IR pass filters and the Heliopan 8125 UV/IR cut filter — based on data from Clive Warren's Infrared Photography FAQ and the Heliopan No. 8125 Digital filter page.
Note that the 89b (R72) has a sliver of transmissivity in the deep visible red just below the 700 nm visible-infrared boundary. Some refer to it as a "dark red" filter because you can see through it to a very limited extent, but that appellation doesn't give the R72 proper credit for blocking nearly all visible light. The "black" 87-series filters, on the other hand, pass no visible light at all.
The R72, 87 and 87c pass progressively less light to the camera's sensor, which itself loses IR sensitivity at ~1200 nm. For all three IR pass filters, peak transmissivities of only 80-90% also contribute to unwelcome light loss in IR work. No wonder, then, that some cameras can handle an R72 but not an 87 or 87c.
The Heliopan 8125 "digital" IR/UV cut filter claims to block unwanted NIR and near UV as well, but the spectrum here shows that most of the longwave NIR still gets through. (That's the kind that theoretically reduces saturation.) I haven't been able to detect any benefit from this filter under normal shooting conditions with several different digital cameras.
![Looking north from Carson Spur, Sierra Nevada, California; R72 IR filter, monopod. Click to see 800x600 version. [C-2020Z]](http://dpfwiw.com/images/headwaters.jpg) |
Looking north from Carson Spur, Sierra Nevada crest, California; handheld C-2020Z withR72 |
On an IR-sensitive camera like my Oly C-2020Z, the R72 typically takes a 5- to 6-stop exposure correction. Handheld shots are often possible in bright sunlight, but solid camera support is still a good idea. With the R72/C-2020Z combination, I've had very good luck with a monopod. The clarity samples shown above were taken by bracing against a post. The upper visible light sample was metered at EV 13.3 and exposed at EV 14.0 to darken the bright haze; the lower R72 version was metered at EV 9.7 and exposed at EV 9.0.
The feature photo at the top of this page and the photo just above were both taken with a Hoya R72, as were many of the stunning IR images dpFWIW contributor Carl Schofield's displays at his Beyond Red... gallery. As illustrated in Carl's much-appreciated samples below, R72 images recorded in color are typically rendered brick red and pale cyan tones. That's as true of my Oly C-20x0Zs as it is of Carl's Nikon CoolPix 950. Truth be told, I've come to like R72 false color scheme.
My experiences with the Hoya R72 match Jay Scott's and Carl
Schofield's: It's truly a joy. (If only my photographs matched Carl's!) To
my mind, the R72 is a great value, a flexible tool, and a forgiving entry into
the fascinating infrared world, and just plain fun. Most digital cameras seem to be able to handle
the R72, but test your camera before you buy.
These black "deep IR" filters pass no visible light to speak of. You can examine the transmission spectra of the original Wratten 87 and 87C gel filters in the chart above.
|
Many digital cameras lack the IR sensitivity to handle 87-series IR filters, especially newer ones. |
That can be true even for cameras that work well enough with an R72 or 89b filter.
![Clouds and trees, Walnut Creek, CA; Wratten 87 IR pass filter, braced monopod, color recording, no grayscale conversion. Click to see 800x600 version. [C-2020Z]](http://dpfwiw.com/images/ir_sky.jpg) |
![Roundtop looking south from Vollmer Peak, Orinda, CA. Wratten 87 IR pass filter, monopod, f/2.2, 1/10 sec. [C-2020Z]](http://dpfwiw.com/images/roundtop.jpg) |
Autumn sky; C-2020Z with 87 on braced monopod |
Looking south from Grizzly Peak to Round Top, Berkeley-Oakland hills, northern California; C-2020Z with 87 on monopod |
The Wratten 87 offers greater contrast and yields nearly pure grayscale output on most digital cameras, even with color recording, as you can see at right and in Carl Schofield's IR filter samples below. Unfortunately, the Wratten 87's additional 2-stop bite out of exposure precludes handholding under most conditions, but I've gotten away with it on very bright days with lots of bracketing for camera shake.
The Heliopan RG780 is an 87 equivalent. I'm unaware of a B+W equivalent for the Wratten 87.
With a 50% cut-off at 850 nm, the pricey black Wratten 87c (B+W 093) operates deeper yet in the NIR. The 87c typically yields blue monochromes like the one shown in the sample table below. (This was not always the case with the 87c on Carl's CoolPix 950, however, as discussed below.) The 87c presumably requires an even greater exposure correction than the 87.
Common IR Filter Comparison Samples |
|||
| Kodak Wratten Filter | 89b | 87 | 87c |
| Hoya Equivalent | R72, RM72 | n/a | n/a |
| B+W Equivalent | 092 | n/a | 093 |
| 50% cut-off mark | 720 nm | 800 nm | 850 nm |
| Nikon CoolPix 950 IR Images, all with auto white balance
and color recording, firmware v. 1.3.
Courtesy Carl Schofield. All rights reserved. |
![Hoya R72 IR filter, auto white balance. Courtesy Carl Shofield; all rights reserved. [CoolPix 950]](http://dpfwiw.com/images/r72-v13.jpg) |
![Wratten 87 IR filter, auto white balance. Courtesy Carl Shofield; all rights reserved. [CoolPix 950]](http://dpfwiw.com/images/87-v13.jpg) |
![B+W 093 (Wratten 87c) IR filter, auto white balance. Courtesy Carl Shofield; all rights reserved. [CoolPix 950]](http://dpfwiw.com/images/87c-v13.jpg) |
I have no personal experience with the Wratten 87C, but it's got to be an even tougher challenge than the 87 on all fronts.
With a 50% transmissivity at ~750 nm, this less common IR filter falls between the 89b/R72 and the 87 series with regard to total light loss. By the numbers, at least, the 88a runs closer to the R72. Some cameras (like the Canon G1) can't handle an 87 but do well with an 88a. The Heliopan RG715 filter is an 88a equivalent.
The short answer is, not much for non-Hoya filters. Kodak Wratten IR filter numbers (89b, 88a, 87, 87c, etc.) tend to go down and Heliopan RG numbers tend to go up with increasing 50% transmissivity wavelengths, but if they mean more than that, it's not at all obvious to me. B+W IR filter numbers (092, 093, etc.) are pretty much meaningless.
The naming system Hoya uses for its IR pass filters is refreshingly rational. The R72 hits 50% transmissivity at 720 nm, just inside the NIR. The RM90 hits 50% transmissivity at 900 nm, and so on. How Hoya R and RM series IR filters differ, I have no idea.
Digital cameras make IR photography easy, fun and affordable, but there are some tricks and pitfalls worth knowing.
On this page...
We'll flesh them out below, but here are the basics of digital IR photography up front:
Carry your IR filters and use them liberally
Know your NIR light sources
Know how to play your camera's ISO vs. noise and resolving power vs. shutter speed trade-offs.
Arrange for adequate camera support
Record in color
Check your work at the scene
When in doubt, bracket like crazy for exposure and camera shake
But first, a very important safety reminder:
|
Never, ever view the sun directly through an IR filter, however black it may appear. |
The transmitted near IR can permanently damage your eyes in a matter of seconds before you know it!
Carry one or more IR filters (at least an R72) at all times and use them liberally — even when you don't "see" any promising subjects. The IR realm is full of surprises, and good IR shots can be impossible to predict from the visible light version. If you limit yourself to landscapes, the traditional IR fare, you'll miss half the fun — and a lot of good IR material.
To see what I mean, take a moment to browse Don Ellis' IR gallery. His striking R72 images of Hong Kong underscore the versatility of the most popular IR filter around. As Don puts it, "... the real image isn't always the obvious one". I'd say that goes double for IR.
Digital IR photos typically record reflected NIR. Emitted NIR is much less commonly encountered but is by no means rare. Modern flash tubes emit enough NIR to be useful with IR filters, but fluorescent lights emit very little. Objects hot enough to glow visibly emit lots of NIR, too, but
|
Objects at room to body temperatures don't glow in the NIR any more than they do in the visible band. |
 |
The sun provides most of the NIR captured in IR photography |
Objects that appear bright in IR images almost always do so because they have high NIR reflectivities. Objects that appear dark in IR images reflect little NIR, and they're more likely to transmit than absorb it. Generally speaking, NIR reflectivity does not follow reflectivity in the visible band. Thanks to the "forbidden" molecular transitions that correspond to NIR wavelengths, NIR emission and absorption are both relatively uncommon.
 |
At this point, you can skip directly to Exposure and Camera Support or read on for more background on IR emissions. |
Emitted NIR comes primarily from objects at temperatures of many hundreds to thousands of Kelvins (°K) — far above body and room temperatures. One way to spot NIR-rich sources is to look for incandescence, the thermal emission of visible light. After a feeble start at temperatures around 500°C (773°K; 932°F), incandescence becomes conspicuous at around 627°C (900°K; 1,160°F), regardless of the material being heated. Since most of the energy radiated at such temperatures falls squarely in the NIR,
That includes glowing coals, electric heater coils, molten metal and glowing lava. Objects heated to 300°C (573°K; 480°F) or above can also glow substantially in the NIR, even when they're not yet incandescent. For example, just before its first dusky red tones appear, an electric heating coil coming up to temperature radiates strongly in the near IR and continues to do so throughout its working temperature range.
The thermal radiation emitted by bodies at room to body temperatures lies in the far IR at wavelengths of 3µ (3,000 nm) or longer — well beyond the reach of the silicon-based digital cameras discussed on this site. (Silicon loses all IR sensitivity at 1.2µ.) Another major hurdle to thermal IR imaging is our atmosphere. While highly transparent at visible wavelengths (0.3-0.7µ) and gloriously transparent at NIR wavelengths (0.7-1.4µ), air is quite opaque at 5-8µ, as shown here. Above 5µ, glass also becomes opaque, so you'll have to resort to mineral lenses that make high-end 35 mm glass lenses look cheap.
So, to "see" body heat in complete darkness, you'll have to give up your silicon-based CCD or CMOS sensor for something truly exotic that operates in the 3-5µ band where air and glass are still reasonably clear. If you have tens of thousands of dollars to spend, you might consider something like the indium antimonide (InSb) focal plane array IR sensor in the military night vision FLIR MilCAM. I don't think we're in Kansas anymore, Toto.
Wein's Law states that a black body at temperature T radiates with peak intensity at wavelength
w0 (nm) = 2.898 x 10^6 / T (°K)
The table below was constructed using Wein's Law. Note that the human body peaks far beyond the 5µ cut-off for atmospheric and glass transparency.
| Black Body Temperatures and Peak Thermal IR Wavelengths | ||||||
|
Temperature |
°F |
°C |
°K |
Peak (nm) |
Peak (µ) |
Band |
|
Solar surface (effective) |
9,941 |
5,505 |
5,778 |
502 |
0.50 |
Green |
|
Average daylight |
9,440 |
5,227 |
5,500 |
527 |
0.53 |
Green |
|
Tungsten lamp |
5,660 |
3,127 |
3,400 |
853 |
0.85 |
Near IR |
|
Hawaiian lava |
2,012 |
1,100 |
1,373 |
2,112 |
2.11 |
IR |
|
Obvious incandescence |
1,160 |
627 |
900 |
3,223 |
3.22 |
IR |
|
Onset of incandescence |
932 |
500 |
773 |
3,751 |
3.75 |
IR |
|
Body (human) |
99 |
37 |
310 |
9,350 |
9.35 |
Far IR |
|
Room |
68 |
20 |
293 |
9,892 |
9.89 |
Far IR |
With most if not all of the visible light cut out of the picture, exposures with IR filters require substantial compensations. On my highly IR-sensitive Oly C-2020Z, a Hoya R72 typically produces 4-6 stops of light loss. I can usually handhold R72 shots on bright sunny days or in other settings with lots of NIR illumination, but I have even better luck with monopod support in such situations. On the same camera, my Tiffen 87 runs about 2 stops darker and nearly always requires at least monopod support.
Some recent cameras are still IR-sensitive enough for handheld IR shots at their widest aperture and highest ISO settings. You may avoid camera shake that way, but you may also end up with suboptimal resolving power, unacceptable noise levels or both. If your camera (like my Oly C-5050Z) has an action-shot program mode favoring short exposures at the expense of wide-open apertures and high ISO, give it a try with handholding and see what you get. You may be able to clean up a good bit of the noise in post-processing with a program like Neat Image, but don't count on that out until you've tried it.
Otherwise, bring along a tripod or some other rock-steady camera support. Remote control triggering would also help. Remember,
With adequate support comes more freedom to choose shutter speed, aperture and ISO according to scene qualities.
Admittedly, adequate camera support can be hard to come by on impromptu IR outings, but the R72 is fairly forgiving, and there's almost always something around to brace against — a tree, a rock, a lamp post, a steady companion. When available support has been less than optimal, bracketing for camera shake has saved many an IR shot for me. Just take 2-4 exposures of everything you shoot. You can't shake all the time.
NIR reflectance patterns |
|
|
|
Visible Light |
Near Infrared |
Keep in mind that objects that appear quite dark at visible wavelengths may be very bright in the near IR. Foliage is the classic example. The opposite may also obtain. Clear skies that look bright blue to you will usually photograph quite dark with an IR filter in place because the atmosphere scatters little NIR. These intensity shifts can lead to unexpected exposure variations in IR shots, especially with spot metering.
For all those reasons, I generally prefer to shoot IR with matrix metering in a program or priority mode, depending on the challenges at hand. I also spend a lot of time previewing with my LCD.
Shoot everything in sight and ask questions later. Digital cameras excel at the instant feedback needed to work through unfamiliar photographic situations. To be safe, load a big memory card and bracket heavily for both exposure and camera shake, especially if handholding. Before shooting in a priority mode, check the exposure settings via your LCD to make sure you haven't exceeded your camera's capabilities.
![Bright IR glare on San Franscisco Bay, Hoya R72. Click to see 800x600 version. [C-2020Z]](http://dpfwiw.com/images/ir_glarem.jpg) |
IR glare off San Francisco Bay |
Atmospheric scatter is seldom a concern in IR work, but unwanted IR reflections may still warrant a polarizer. I should have used one to suppress the bright sunlight reflecting off the bay in the R72 photo of San Francisco at right. I also can think of a few snowy IR scenes where an upside-down GND might have helped, but the naturally dark skies in IR photos generally make life easier on the excess contrast front. At altitude, your IR shots might benefit from a UV-cutting haze filter.
As always, keep an eye out for flare and vignetting when stacking filters.
Here's a counter-intuitive one for you: Color isn't even a meaningful concept in the NIR band, but color recording is still your best bet for high-quality IR work. As in B&W work, and for all the same reasons, you'll have more control and more options in post-processing with color recording, and you won't have burned bridges with a simplistic in-camera grayscale conversion algorithm.
With color recording, early digital cameras like the Oly C-2020Z and the Nikon CoolPix 950 rendered R72 images in a characteristic and I think rather pleasing brick-and-cyan false color scheme. Wratten 87 series images usually came out as pure grayscales. These false color schemes are illustrated above and discussed in more detail both above and below. Later cameras with more sophisticated Bayer color interpolation and white balance algorithms tended to render Wratten 87c color recordings as blue monochromes and R72s as magenta monochromes garish enough to make anyone wince — at which point the hunt was on for ways to manipulate the false colors, both in-camera and at post-processing.
If your camera's IR false colors don't suit, you have several alternatives, depending on available camera features, time, skill and software tools:
Color recording with grayscale conversion in post-processing
Color recording with color manipulation (red-blue channel swapping, admixture of channels from IR and visible light versions of the same image, and so on) in post-processing
Color recording with manipulated in-camera white balance (keep showing cards of different colors to your camera's "show me something neutral" custom white balance function until you get a false-color scheme you like)
RAW color recording with white balance manipulation at conversion to an RGB image.
Grayscale (B&W) recording mode
Many experienced digital IR photographers greatly prefer the power and flexibility of the first approach, but some assembly is required. B&W recording is certainly more convenient, but you'll need to stay on top of the recording mode in effect to avoid mishaps between IR sessions.
Personally, I've come to like the look of R72 color images. My cameras offer B&W recording, but I always record IR images in color.
The short answer: Stick with auto-focus (AF) and you won't have to worry about wavelength-related focus shifts.
Depending on the spectral characteristics of your lens (achromat, apochromat, etc.), IR images are theoretically subject to a wavelength-related focus shift. Typically, IR light comes to a focus just past the focal plane, which has of course been positioned for visible light. If NIR is the only light coming in, AF should be able to adjust accordingly — provided AF has enough light to do its magic. IR filters don't seem to interfere with AF accuracy on most digital cameras — certainly not on the Oly C-2020Z and the Nikon CoolPix 950. Focus shift can be problematic in IR film work, even with AF, but on the digital side, it's also largely absorbed within the generous depth of field typical of consumer-grade digital cameras.
If you have trouble focusing your IR filters, try increasing depth of field by stopping down the aperture and reducing subject magnification as best you can.
The distinctive digital IR look is a many-layered thing. This page will explore its origins, which will in turn suggest ways to manipulate it.
On this page...
|
If you're not in the mood for speculation, skip ahead to a discussion of IR contamination. |
The top layer is the source of near IR (NIR) illumination, most often the sun. The intensity of sunlight may peak at green wavelengths, but sunlight is loaded with NIR as well — particularly at the shortest NIR wavelengths. The same is true of incandescent illumination. Conventional flash units produce a different spectrum, as do IR flashes and the LED-based IR illuminators available today. Whatever the source, the balance between longwave and shortwave components will have an impact on the way your IR photos look.
Next come the NIR reflectivities of the elements in the scene. In the bulleted paragraphs below, the links lead to thumbnails of images that illustrate the brightness relationships discussed. In each case, the source is assumed to be solar NIR.
Deciduous tree leaves and grasses are almost always very bright in the NIR. Although made predominantly of NIR-transparent materials, leaves reflect NIR very efficiently because their complex internal air spaces offer many opportunities for shallow-angle internal reflections that eventually bounce the NIR out again. Leaves lose their NIR reflectivity when soaked in water and pressed to drive out the air. In the visible band, tiny airspaces work the same magic in fallen snow and fluffy soap suds, both of which start out bright white but approach transparency as trapped air escapes. Plant temperatures and pigments like chlorophyll have nothing to do with leaf reflectivity at NIR wavelengths.
Leafless deciduous branches are also quite bright.
Conifer needles tend to be less bright than deciduous leaves for reasons I have yet to divine, but internal structure differences are the prime suspect.
Clouds are very bright and the clear sky dark because condensed water droplets scatter all NIR wavelengths very efficiently, but air molecules hardly scatter them at all (and then only at the shortest NIR wavelengths).
Water surfaces tend to be dark in the NIR, mainly for want of bright skylight to reflect.
![Traffic sign, no filter. Click to see 800x600 version. [C-2020Z]](http://dpfwiw.com/images/no_turn.jpg) |
![Traffic sign, Hoya R72. Click to see 800x600 version. [C-2020Z]](http://dpfwiw.com/images/no_turn-r72.jpg) |
Visible |
NIR (R72) |
I had a lot of fun on this IR stroll through Denver's Washington Park, the location for many of the images in this article. Checking out the NIR world though the LCD of a digital camera with an IR filter mounted is a good way to learn to pre-visualize IR shots. It's also full of surprises.
Next in line among the many layers contributing to IR look come several layers related to the camera itself:
the external IR filter used,
the Bayer pattern color filters applied to the CCD,
the internal IR cut filter applied to the CCD,
the CCD itself, and
the camera's firmware—particularly the color interpolation and white balance algorithms in effect.
Each camera layer puts its own stamp on the final IR image.
Solid information regarding the last 4 camera layers listed above is hard to come by. Silicon-based CCD and CMOS sensors are equally sensitive at visible and NIR wavelengths but abruptly lose all sensitivity at around 1,200 nm. The CCD thus puts a fixed limit on NIR input to the image at the long end of the spectrum, while the external IR filter imposes a variable limit on NIR input at the short end.
The proprietary internal IR cut filters sensor manufacturers apply to their digital camera products remain shrouded in mystery. I have yet to see a transmission spectrum for one, but they clearly vary widely and seemingly arbitrarily from camera to camera, as chronicled above. The inner workings of camera firmwares also seem to be hush-hush, but we manage to peek under the hood below.
The sections that follow focus on 3 critical camera-related layers,
the external IR filter used,
the Bayer pattern color filters applied to the CCD,
the white balance algorithm used,
and what we can surmise of their intertwined contributions to the final IR image.
Color-mode digital IR images have a look and feel that varies with the IR filter and to a lesser extent with the camera used. The popular dark red R72 filter typically produces images with either a brick red and pale cyan or a red and magenta color scheme, depending on the camera, while images made with the deeper 87 filter tend to come out as grayscales or nearly so, regardless of the camera used. Representative color-mode digital IR samples are shown again in the table below.
Color is, of course, an attribute of visible light alone. Since color is meaningless at NIR wavelengths, any color found in an NIR image is by definition false color, gray included. A digital camera is nevertheless compelled by its firmware to do something with the sensor data an IR filter generates. The samples above are representative of what camera firmwares tend to come up with when confronted with NIR scenes recorded in color.
How do these false colors arise? All available clues point to white balance algorithms and the NIR spectral properties of the Bayer pattern color filters covering the CCD sensels in single-CCD color digital cameras. The shortest NIR wavelengths (in the 700-770 nm band) appear to drive most of the false color rendering, as we'll see.
Based on several different lines of evidence, I've concluded that one must simultaneously consider what goes on in 2 functionally different NIR bands in order to understand the false colors seen in digital RGB images taken with IR filters:
700-770 nm, hereafter referred to as shortwave NIR
770-1100 nm, hereafter referred to as longwave NIR
Longwave NIR (770-1100 nm) stimulates all sensels equally because Bayer pattern filter dyes are uniformly transparent at those wavelengths for quantum mechanical reasons elaborated in the sidebar. Thus, longwave IR primarily affects false-color saturations.
Shortwave NIR (700-770 nm), on the other hand, stimulates red sensels quite a bit, green sensels a little and blue sensels minimally if at all. Shortwave NIR thus drives the false colors, especially with shallow filters like the R72 (50% cut-off at 720 nm). The red tinge typical of R72 skies bears this out: Shortwave NIR is heavily over-represented in what little NIR the atmosphere manages to scatter because scattering efficiency by air molecules falls off inversely with the 4th power of wavelength.
If you block most of the shortwave NIR with a deeper Wratten 87 IR filter (50% cut-off at 800 nm), you get a near-perfect grayscale image, as you'd expect with equal stimulation of all 3 sensel types by the remaining longwave NIR.
Things get even more interesting with the Wratten 87c (50% cut-off at 850 nm), which in many digital cameras produces blue monochromes rather than the grayscales one might expect. I believe that this reflects a slight blue bias built into most auto white balance algorithms to counter the anti-blue bias carried by NIR contamination in visible light digital images.
|
At this point, you can skip directly to a discussion of IR contamination in visible light images, or you can brace yourself for a slog through the arguments supporting the theory above. |
In an earlier version of this section, I wondered out loud about why digital IR images tend to be monochromatic, especially with deeper IR filters like the 87 series. Andrew Fong, a Ph.D. analytical chemist whose dissertation dealt with the use of NIR radiation in chemical analysis, e-mailed me this compelling answer:
In my work, we generally use wavelengths greater than about 1200 nm to around 3000 nm, but I have some knowledge of the NIR band around 1100-770 nm.
The short answer is that the CCD mask filters are all nearly equally transparent to short NIR radiation [at 770-1100 nm]. The blue filters absorb some of the NIR radiation as a spill-over of their red-absorbing properties.
Basically, most substances are fairly transparent to this short-wavelength NIR radiation [at 770-1100 nm]. Some larger dye molecules can be made to absorb there (an electronic type absorption). However, this energy region does not happen to correspond with most molecular vibrations or electronic absorptions. Most molecular vibrations correspond with the Mid-IR [> 3000 nm]. Overtones and combinations of these Mid-IR vibrations occur between 3000-1200 nm. By the time you get to 1100-770 nm, these overtones and combinations of molecular vibrations become very weak (they are quantum-mechanically forbidden). Therefore, there aren't many compounds which absorb any light at all between 1100-770 nm unless they are dyes specifically designed to absorb there. (Water vapor does have a very weak absorption band somewhere around 900 nm).
Therefore, all of the color filter types on the CCD probably pass almost all of the short NIR radiation [at 770-1100 nm]. It is possible that the blue CCD mask filters absorb some of the light at the shortest end of the NIR band (an electronic absorption tail).
—Andrew Fong, Memphis, TN USA
[Text in square brackets added by editor for clarity.]
With no suitable molecular energy gaps to fall into, longwave NIR photons at 770-1100 nm can be transmitted or reflected, but they're seldom absorbed. Shallow angles of incidence will promote reflection off NIR-transparent materials, just as they produce strong reflections off clear glass in the visible band. The complex air spaces found within deciduous leaves and fallen snow set up strong NIR reflections in just this manner. In the absence of reflection, longwave NIR transmission will dominate. Note that the forbidden NIR absorptions start at 770 nm, not at the visible-IR boundary at 700 nm. Allowed shortwave NIR absorptions in the 700-770 nm band will become important below.
Andrew's explanation jibes with dpFWIW contributor Jay Scott's early observation that the red CCD filters in his Oly C-2020Z appear to pass shortwave NIR preferentially, while the blue CCD filters preferentially pass longwave NIR. It also jibes with Oly C-2020Z CCD spectral response plots showing ~70%, ~10% and ~0% respective sensitivities for the red, green and blue sensors at 700 nm, the visible-IR boundary. I have no hard data for shortwave NIR absorptions at 700-770 nm, but these plots are also consistent with Jay's observations.
Now we're ready tackle the false colors produced by some of the more popular IR filters. What goes on in the undocumented 700-770 nm seems to be the key.
The R72 produces images rendered in brick red and pale cyan tones by some cameras and in bright red to magenta colors by others. To my mind, the brick and cyan R72 pattern suggests an overall excess of red sensel stimulation and a blue sensel stimulation deficit, presumably due to the red filter's relatively high transmissivity and the blue filter's electronic absorption tail at the shortest NIR wavelengths.
I don't have a ready explanation for the red and magenta R72 color scheme, which has become more and more common in recent cameras. At one time I wondered if it might occur with CCDs sporting CYGM (cyan, yellow, green, magenta) rather than GRGB (green, red, green, blue) Bayer pattern filters. However, the CYGM Canon G1 renders R72s in a purple-and-cyan color scheme similar to the brick-and-cyan scheme my GRGB Oly C-2020Z produces. See Don Ellis' IR gallery for some G1/R72 samples. Todd Walker's Canon IS Pro90 R72 samples resemble Don's G1 R72s very closely, as you might expect from 2 cameras built around the same 3.34MP CYMG CCD.
Since Bayer pattern filters appear to have some differential effect on shortwave NIR (the 700-770 nm band), those must be the wavelengths primarily responsible for the false colors typical of images taken with shallower IR filters (like the R72 and the 88a) with 50% transmissivities in the 720-750 nm range. I believe that the sky takes on a reddish tone in the R72/89b sample above because the atmosphere scatters less longwave than shortwave NIR, which in turn preferentially stimulates the red sensels. Foliage and clouds reflect all NIR wavelengths equally but acquire a cyan cast from a white balance algorithm trying to compensate for an overall excess of red sensel stimulation.
Beyond 770 nm, forbidden absorptions imply grayscale output from equally transparent red, green and blue filters, and grayscale is precisely what the Wratten 87 filter delivers, regardless of the camera. Why? Because the 87's 50% cutoff at 800 nm blocks both visible light and most of the color-generating shortwave NIR below 770 nm. Mild red and green sensel stimulation by the small amount of shortwave NIR transmitted by the 87 offsets an apparent blue bias built into typical auto white balance algorithms for reasons I'm about to propose.
The blue monochromes typical of 87c filters provide important clues to the IR false color puzzle.
With a 50% transmissivity deep in the NIR at 850 nm, the 87c absorbs virtually all of the color-generating shortwave NIR. These are the same NIR wavelengths blocked by the blue filter's electronic absorption tail in visible light work. The 87c thus deprives the red and green sensels of the shortwave NIR photons normally unavailable only to the blue sensels. A white balance algorithm conscious of NIR contamination and preprogrammed to boost blue sensels a bit to compensate for their normally reduced NIR stimulation would automatically add a touch of blue to the pure grayscale the 87c should otherwise have produced.
If you're still skeptical of the role white balance plays in IR false colors, consider this data point: Under the original v. 1.1 firmware in his Nikon CoolPix 950, Carl Schofield routinely got pure grayscale images from his Wratten 87 and B+W 093 (Wratten 87c equivalent) filters with color recording. But when his camera came back from a factory repair with an unexpected upgrade to v. 1.3 firmware, Carl was surprised to find his previously grayscale B+W 093 (87c) images coming out as blue monochromes instead of grayscales, as shown above. Interestingly, the firmware change had no visible effect on his visible light, R72 or Wratten 87 images. Whatever else Nikon might have updated between v. 1.1 and v. 1.3, white balance handling was one of the acknowledged firmware changes.
My Oly C-2020Z handles R72 and Wratten 87 color images just like Carl's CoolPix 950. I have no Wratten 87c to test, but when I stack a hot mirror (IR cut) filter on top of my Wratten 87 to block out all remaining shortwave NIR, I get blue monochromes, too.
To mind, these observations point to a strong white balance input.
Given the camera manufacturers' reticence regarding firmware and internal IR cut filter details, we may never know for sure, but the evidence pieced together so far is fairly compelling: White balance and color filter spectral properties in the NIR both play key roles in the false colors and monochromes that appear in digital IR images recorded in color.
If anyone out there has solid information along these lines, please drop me an e-mail at dpfwiw@cliffshade.com.
Unfortunately, IR sensitivity isn't all good news for digital photographers. For those with more IR-sensitive cameras, the otherworldliness of many IR images raises a disturbing question:
If my camera's IR-sensitive enough to take decent IR photographs, and if my NIR and visible light photos differ that much, shouldn't I be worried about IR contamination in my visible light work?
The welcome short answer seems to be "seldom if ever", but just how we've managed to get off the hook on this score isn't all that clear.
On this page...
Detectable IR contamination turns out to be quite rare in digital
photographs,
even with IR-sensitive cameras like my Oly C-2020Z.
Most convincing instances of IR contamination involve objects at or near incandescent temperatures. Such objects emit thermal IR in the NIR band. To see an example of IR contamination from a hot object, visit Peter iNova's dpreview.com infrared tutorial and scroll down to the first image of the burning gas-fired radiant heater. The bluish sheen on its very hot heat reflector is no doubt thermal NIR. You can count on IR contamination with any object that hot, but how often do they pop up in your photographs?
In setting up the Filter Test — Color Bias and Saturation in another dpFWIW article, I managed to create a related instance involving a close-up under hot, close-range incandescent lighting. Here, the contaminating NIR was emitted by a 60W bulb and reflected to the camera by a room temperature subject over a lamp-to-camera distance of ~20 inches.
Situations like these are both rare and foreseeable. I have yet to see a compelling example of IR contamination in an outdoor shot containing objects at ordinary temperatures. Theoretically, one would expect IR contamination to be most apparent in visible light digital photos of visibly dark but IR-bright objects like foliage, but accurate renditions of foliage don't seem to be a problem for digital cameras.
Many have tried to pin the common purple fringing artifact on IR contamination, but I their arguments unconvincing. Purple fringing is primarily a high-order lens aberration.
So, go ahead and worry about IR contamination when you're shooting very hot objects or shooting very near such objects. Otherwise, forget about it.
Sunlight, the most commonly encountered NIR source, contains both shortwave (700-770 nm) and longwave (770-1100 nm) NIR in abundance. Sunlit objects should be subject to IR contamination, particularly visibly dark NIR-bright objects like foliage.
Since Bayer pattern filters are equally transparent to longwave NIR, longwave contamination should add white to affected areas of the image. That would desaturate affected colors and might also contribute to overexposure and blooming, particularly in areas already close to overexposure. My Heliopan 8125 "Digital" UV/IR cut filter strongly attenuates longwave NIR, but it doesn't visibly improve color saturation in my outdoor work. Nor does it tone down blown-out leaf highlights. The 8125 did improve saturation in the Filter Test — Color Bias and Saturation described elsewhere on dpFWIW, but that was a very special circumstance unlikely to be encountered in the field.
Shortwave NIR contamination should cause color shifts toward the red and might contribute to white balance failures as well. Red Bayer pattern filters are most transparent to shortwave NIR; green filters are slightly transparent and blue filters least transparent, at least in my C-2020Z.
I used to blame shortwave IR contamination for my camera's tendency to overexpose red flowers in bright sunlight, but now I'm not so sure. My Heliopan 8125 "Digital" UV/IR cut filter does nothing to help these blown-out reds. Then again, flower colors can be tough, and the 8125's much less effective against shortwave than longwave NIR. The jury's still out on this one.
To get a handle on potential IR contamination, dpFWIW contributor Jay Scott performed a clever and interesting "black object test" with his Oly C-2020Z and a Hoya R72 filter:
I ran some tests last night, and now I'm convinced that it's sometimes a serious problem.
The most common color problem I've seen with my camera is for black cloth to occasionally be rendered blue, especially indoors under incandescent light. I reasoned that, first, any big color shift in a black object that doesn't reflect much visible light almost has to be due to unwanted sensitivity to invisible light, and second, if there is unwanted sensitivity, a black object is what will show it most. And incandescent lights are very bright in the infrared. So I went around the house and collected (1) an object that was black and photographed black, (2) an object that was black and photographed blue, and (3) an object that was black and photographed blue under incandescent light and nearly black with xenon flash. Object (1) turned out to be infrared dark. Objects (2) and (3) turned out to be infrared bright. Flash is infrared-bright, but it is much bluer than incandescent light, so I infer that (3) is only bright at longer infrared wavelengths, while (2) is bright over a wider spectrum. Fluorescent lights are dark in the infrared, so I put them under fluorescent light and all three photographed black. I found a bunch of other black objects that photographed black, and all of them turned out to be infrared dark.
I'm convinced. My images are contaminated with infrared light, and in the worst case it makes a big difference. I still haven't done tests to see how big a difference it makes in typical cases. I suspect this may be what makes purple flowers so often rendered as blue.
Suddenly I'm in the market for a hot mirror....
It hardly ever pays to disagree with Jay, but if it's going to take a hot mirror to cure IR contamination, the cure could well turn out to be worse than the disease. To see a hot mirror in action, check out this filter test.
|
If you lay awake at night worrying about IR contamination, press on to learn about two different filters you might use to block it. Otherwise, skip ahead to the References and Links and visit some of the IR galleries listed there. |
External hot mirror filters are the primary line of defense in what I now see as a largely theoretical battle against IR contamination. Their dichroic (dielectric) coatings reflect rather than absorb IR. Most consumer-grade cameras come with internal hot mirrors already installed on their sensors — the internal IR cut filters discussed so many times before. A typical hot mirror transmission spectrum can be viewed here.
The cheap no-name hot mirror I use with my Tiffen 18A UV pass filter blocks a significant amount of visible red light along with the IR. The result is an ugly greenish cast in visible light images, as illustrated here. In UV photography, that's not a problem, but this hot mirror is no viable cure for IR contamination in visible light work. Most of the hot mirrors for sale at B&H Photo are very expensive, but they cause visible artifacts as well. The professional photographers who buy them work with digital backs for 35 mm and and medium-format cameras, and they may well need them, but consumer-grade digital camera users seldom if ever do.
One of the more promising IR cut filters would seem to be the multicoated Heliopan No. 8125 Digital, which claims to block both UV and IR while freely passing all visible wavelengths. As its complete transmission spectrum clearly shows, the 8125 strongly attenuates longwave NIR and UV-B and UV-C, but it passes a good bit of shortwave NIR and some near UV-A as well. That makes the 8125 a very leaky defense against IR and UV contamination. In fact, it's about as ineffective against shortwave NIR as commonly available UV cut filters are against UV-A.
Several months of casual shooting with the 8125 confirm the story told by the spectrum. I have yet to see an outdoor benefit on the IR or UV side — even with sun-drenched foliage, one of likeliest IR contamination scenarios, as illustrated in the sample images below.
Heliopan 8125 Outdoor Test |
|
| Wratten 87 IR pass filter image documenting the test scene NIR content. The 87 blocks all visible light and some shortwave NIR as well. As you can see, there's no shortage of longwave NIR in this sunny day at the park. | ![Wratten 87; 800x600 [C-2020Z]](http://dpfwiw.com/images/test87-m.jpg) |
| Control image, no filter. Note the visibly dark but very IR-bright deciduous trees. If the 8125 were effective against IR contamination, that's where you'd be most likely to see it. | ![No filter; 800x600 [C-2020Z]](http://dpfwiw.com/images/testbare-m.jpg) |
| Heliopan 8125 "Digital" UV/IR cut filter produces no visible effect, despite its strong longwave NIR attenuation. Its UV attenuation plays no role here. | ![Heliopan 8125 "Digital" UV/IR cut filter; 800x600 [C-2020Z]](http://dpfwiw.com/images/test8125-m.jpg) |
Nor did the 8125 help with the substantial red over-saturation my then C-2000Z tended to impart on reddish flowers taken in bright sunlight with auto white balance.
To be fair the 8125 did manage to improve color saturation substantially in the Filter Test — Color Bias and Saturation described elsewhere on dpFWIW, but I don't expect to encounter a similar situation in the field anytime soon. If NIR contamination were truly a practical concern in outdoor digital visible light photography, the 8125's strong longwave NIR attenuation should have improved color saturation in the test shots above. I don't see that it did.
So, in the absence of a detectable benefit in worst-case outdoor photos like the test images above on an unusually IR-sensitive camera like my Oly C-2020Z, I conclude that
and that the 8125's cost/benefit ratio is far above recommendable levels. (In 2Q 2000, mine cost $79.95 on special phone order from B&H Photo.)
If anyone has solid evidence to prove me wrong here, I'd love to see it.
The internet is loaded with material pertinent to IR photography, but the best sites are the IR galleries.
On this page...
B&H Photo — if they don't have it, or can't get it for you, I'd seriously consider giving up.
eBay — the IR aficionados' best bet for legendary IR cameras like the Oly C-2020Z and the Nikon CoolPix 950.
Neat Image — sophisticated yet affordable noise reduction software with few equals.
Integrated Design Tools - High speed digital technology, cutting edge stuff.
The Filter Connection — a good source for IR and other exotic filters, filter information and filter-related camera accessories, including lens hoods and multicoated filter cleaners. Best of all, you even can discuss your filter purchases with a real live knowledgeable human!
