infrared issue18 ctdp 2015

17

Upload: mariosapereira

Post on 13-Jul-2016

5 views

Category:

Documents


0 download

DESCRIPTION

Infrared Issue18

TRANSCRIPT

Page 1: Infrared Issue18 CtDP 2015
Page 2: Infrared Issue18 CtDP 2015

Infrared photos show the world in a completely newlight. Digital sensors or special film can be used tocapture image data that lies beyond the limits of thevisible spectrum, where you can find highly unusualcolor and contrast effects. This article explains whatinfrared light is and introduces the techniques requiredto capture and process your own infrared images.

Canon EOS Rebel T3i/600D | Non-specific conversion/Astronomik 742nm clip-in filter | ISO100 | f7.1 | 1/500s | Channel swap in Difference blend mode

Ralph Altmann

Infrared Photography

Page 3: Infrared Issue18 CtDP 2015

We live in the midst of a flood of invisiblelight that is in fact just as bright as the

familiar visible spectrum, which containselectromagnetic waves with wavelengths thatrange from 400 nanometers (blue) to 700nanometers (red). Shorter wavelengths belongto the ultraviolet spectrum and are responsiblefor tanning our skin in the sun, while weperceive longer wavelengths in the form ofwarmth. Wavelengths that are slightly longerthan visible red generate no perceptiblewarmth and are referred to as Near Infrared(NIR) waves. The electromagnetic wavesinvolved in infrared photography are the‘A-band’ NIR waves with wavelengths of780-1450 nanometers (nm).

The Sun is the main natural source ofinfrared light, although artificial lights alsoproduce some infrared frequencies as anunplanned side effect. Specially designedinfrared spotlights and flash units thatproduce only invisible radiation can be usedto capture nighttime photographs – ofnocturnal animals, for example. Livingorganisms and even comparatively coldobjects produce infrared radiation too, but ofsuch long wavelengths (10,000 nm and more)that you would need to use specially madesemiconductor-based sensors to captureimages of them using a process calledthermal imaging.

The silicon-based chips that digital camerasensors are based on can only capture a verysmall part of the infrared spectrum, but eventhese frequencies are usually suppressed fortechnical reasons. If you have an old digitalcamera lying around (and you don’t mindpossibly breaking it), you can convert it into a specialized infrared camera. You can then shoot handheld using exposure timesvery similar to those you would use undereveryday shooting conditions.

But you don’t have to dive in anddismantle your camera to experiment withinfrared effects. With the right filter, a tripod

and long exposure times, virtually any digitalcamera can be used to capture infraredimages. But before we go into the technicaldetails, let’s take a look at what infrared (IR)photography is and what makes it sofascinating.

The Wood Effect

The world looks very different in infrared light.Leaves and grass reflect IR light almostcompletely and appear very bright (tendingtoward white). This effect was discovered byUS physicist Robert Williams Wood almost 100years ago.

IR light is not, however, reflected by thechlorophyll in leaves, which is transparent toIR frequencies, but rather by the lower layersof spongy parenchyma that make up the fleshof a leaf. This is why patterns on the surface ofleaves disappear almost completely in IRimages. Many other pigments also behave asif they are non-existent in IR light: blackT-shirts appear white and sunglasses becometransparent. Skin takes on a smooth, waxylook and everyday blemishes disappear. IRphotos also unmask dyed hair, but thepurported ability to look through people’sclothing using IR light is a myth and wasprobably only possible in the case of someearly nylon fabrics.

The Wood Effect drastically alters thedistribution of contrast in landscape scenes.Regardless of how dark they are in normallight, plant-based greens appear bright, soleaves and evergreen needles contraststrongly with stems, branches and other non-green material, giving IR images of thick, darkforests a translucent, finely detailed look. Theunusual mood in IR images of forests isunderscored by the extremely dark blue tonesthe technique produces. Blue skies becomealmost black without the use of a polarizingfilter and clouds appear especially bright andthree-dimensional. Mist and fog disappear

completely. Many IR photos appear to havebeen created by inspired painters rather thanby the sensor in a digital camera. Our imageof the Muskau Palace in Görlitz, Germany onpage 108 is a great example. These types ofeffects simply cannot be simulated digitally.

What Color is Infrared?

The fact that we perceive green as green, redas red and blue as blue actually has nothing

110 c’t Digital Photography 18 (2015)

Infrared Photography | Introduction

NormalInfraredwith a 720nm filter

400 500 600 700 800 900 1000

Camera light sensitivity Wavelengths reflected by healthy vegetation

~750 nm 970 nm

550 nm

500 nm 675 nm

In black and white, infrared green tones appear darker than normal

Viewed in infrared light, the green of plants and trees appears muchbrighter than it does in visible light

What is IR Photography? 110

Filters and Their Effects 114

Modifying Your Camera 117

Post-processing 122

In this article

Page 4: Infrared Issue18 CtDP 2015

to do with the physical attributes of lightwaves. Color is purely interpretive and is theresult of a trick played by the human brainthat enables us to distinguish betweenvarious wavelengths of light when we viewthem simultaneously. The ‘hardware’ requiredto distinguish between different wavelengthsare the rods, cones and ganglions that makeup the human retina. The different types ofcells are sensitive to different portions of the visible spectrum. To perform the samefunction in a digital camera, an array ofdifferent colored microlenses is positioned in front of the sensor, thus creatingphotoreceptors with differing maximumsensitivities of 450 nm (blue), 550 nm (green)and 630 nm (red). Mixtures of these threeprimary colors are then used to recreate allother visible colors.

The same is true for infrared frequencies,but we have no innate mechanism for‘encoding’ these strange ‘colors’. What we cando, however, is redistribute the colors we canperceive among the wavelengths we capture.This is simultaneously the charm and thechallenge of color infrared photography. Inprinciple, it is nothing more or less than aspecific type of pseudocolor editing. Photosare often recolored for documentarypurposes, but ours is a much more artisticapproach. Successful infrared photos don’t

completely change the colors in an image, butinstead skew familiar colors to give us a newview of commonplace subjects. This is a greattechnique for producing images with afairy-tale feel and giving everyday objects amagical veneer.

Beyond Color

The question of which colors to assign towhich wavelengths is only relevant if yourcamera is capable of differentiating betweenthe wavelengths in question. Thetranslucency of all three colors of Bayerpattern microfilters is very similar abovewavelengths of about 800 nm, which meansthat the photoreceptors in the camera’ssensor are all exposed to approximately thesame degree. The result is three colorchannels in which the same image is capturedwith only slight differences in intensity – inother words, the differences in wavelengthare no longer relevant and the capturedimage is monochrome

If captured using a daylight white balancesetting, such an image will consist largely ofred tones, whereas using Auto white balanceusually produces blue or green tones. Thecloser a frequency lies to the visible spectrum,the more the red-toned photoreceptors areused to capture it. Most Auto white balance

mechanisms are simply not up to the job ofdifferentiating between major color shifts, butmanual white balance (applied either duringshooting or later during RAW conversion)often does the trick. However, the result is stilla weakly toned monochrome image. Furthermanual toning can produce interestingresults, but most photographers prefer tomaximize the effect of the increased contrastbetween the sky, clouds and vegetation byconverting such images to pure black andwhite. Monochrome IR images are oftenextremely effective but they don’t have thesurreal, artistic feel of their color counterparts.

All Plants are PinkIf humans were able to see the IR frequenciesthat lie close to the visible spectrum, wewould most likely perceive all vegetation asbright red or pink. Normally, plants appeargreen because the color of chlorophyl has awavelength of about 550 nm (a yellowishgreen) and doesn’t absorb as much light asmaterial with neighboring wavelengths.However, green leaves don’t reflect muchmore than 10 per cent of the incident (visible)light. This is why it is so difficult to capturestrong greens in vegetation located in theshade or under an overcast sky – without a lotof post-processing, leaves simply look gray

111c’t Digital Photography 18 (2015)

Basics | Infrared Photography

Which colors you use to tone theIR-reflective plants and trees in your imagesis largely a matter of taste and depends onthe effect you wish to achieve as well as thedegree to which you are prepared to distortreality

Canon EOS 600D | Neutral modification/HeliopanRG 645nm filter | ISO100 | f7.1 | 1/320s | Post-processed in Photoshop

Page 5: Infrared Issue18 CtDP 2015

and matt. It is only at wavelengths of 750 nmand more (i.e., just beyond the visiblespectrum) that the Wood Effect cuts in andvegetation begins to reflect more light. Atthese kinds of wavelengths, reflection is up tofive times more intense than it is for visiblegreens, producing really bright image areaswhere chlorophyl is present.

Photographed in this type of light, plantsappear much brighter than blue skies, whichis the opposite effect to that produced invisible light. If you check out the individualcolor channels for a conventional landscapeimage, you will find that blue sky is brighterthan vegetation not just in the blue channelbut in fact in all three. In order to produce

deep blue tropical-style skies, you will usuallyneed to use a polarizing filter or apply anappropriate effect during post-processing.

Captured in infrared light, skies alwaysappear dark and provide strong contrast withthe cloud, producing highly effectivemonochrome images. If, however, we wish toshoot IR color images that show blue sky, weare confronted with the problem thatunprocessed sky in an infrared photo has areddish brown color. If we make the sky blueusing a manual white balance adjustment, therest of the image ends up looking blue too,because the translucency characteristics ofthe Bayer pattern microfilters are ‘wrong’ forcapturing infrared images. Relatively short NIR

wavelengths activate the red photoreceptorsand longer wavelengths the green ones. Thevery longest ones activate the bluephotoreceptors at the right-hand end of theNIR curve (see the diagram on page 114). In IRlight, the weak radiation of the blue sky endsup being captured in the red RGB channelinstead of in the blue channel.

How to Rescue Blue Sky

The obvious solution to this dilemma is toswap the red and blue channels, thusassigning shorter wavelengths to the bluechannel and longer ones to red, as is the casefor the visible spectrum. This approach turns

Infrared Photography | Basics

Once we had applied a manual white balance to the original blue-toned RAW image (left), we discoveredthat almost all the color nuances had disappeared (center). The best treatment for an image like this is toconvert it to pure black and white and increase contrast to emphasize the details (right).

Sony DSC-F828 | Nightshot mode/B+W 093 830nm filter | ISO64 | f2.5 | 1/20s

The unusual brightness of all types ofvegetation often creates an illusion of snowor frost. This ‘winter’ view was actuallycaptured in mid-summer at Muskau Park onthe border between Germany and Poland.

Olympus PEN E-PL1 | M.ZUIKO 12mm f/2.0 |700nm conversion | ISO100 | f8 | 1/160s |

Panorama constructed from three source images

Page 6: Infrared Issue18 CtDP 2015

the sky blue and leaves vegetation red.Unfortunately, the frequencies that thedifferent colored microfilters allow throughoverlap to a fairly high degree. Redphotoreceptors are sensitive to visible redlight and frequencies from all parts of the NIRspectrum, while green photoreceptors onlybecome active at the border between thevisible and IR spectra (although they are alsosensitive to all NIR frequencies). Only the bluephotoreceptors display a degree of selectivity;they are only sensitive to IR frequencies.

The exact overlap frequencies vary fromcamera to camera. Typically (as shown in the

graphs overleaf), the blue channel issensitive to IR frequencies above 820 nm,the green channel to mixed sources ataround 720 nm and the red channel to theentire available spectrum (the lower limit isdetermined by the transmission curve of theIR filter you use and the upper limit by the IRsensitivity of your camera’s sensor). The IRlight reflected by vegetation can thereforebe found in all three channels – a situationthat would produce monochromatic colorsin a conventional RGB image.

Swapping channels doesn’t alter the basicsituation but does make the red channel the

‘infrared’ channel in which pure IR light iscaptured, while the blue channel becomessensitive to all frequencies. This means thatthe original reddish-brown image is recoloredin turquoise and blue tones. This effect can becountered using the Levels and White Balancetools and the result is an image with blue skyand bright, pale vegetation. This is the lookfound in most color IR images on the Web.You can then darken the sky and alter thecolor of the vegetation digitally. Manyphotographers like to use yellow for this. See page 122 for more details on post-processing.

113c’t Digital Photography 18 (2015)

Swapping the red and blue channels andadjusting contrast in the resulting

channels enabled us to produce a colorimage from the reddish brown original. To

capture such an image, you need an IRfilter with a relatively low threshold

wavelength that allows some visible redwavelengths to pass.

Olympus PEN E-PL1 | M.ZUIKO12mm f/2.0 | 700nm conversion | ISO100 | f8 | 1/160s

In this image, we made the pinks moreintense by using a daylight white balancesetting and automatic contrast. Weproduced the blue in the sky using thechannel swap technique and the PhotoshopDifference blend mode.

Olympus PEN E-PL1 | M.ZUIKO 12mm f/2.0 | 700nm conversion | ISO100 | f8 | 1/160s

Original Auto contrast

R B channel swap

Page 7: Infrared Issue18 CtDP 2015

Which of the myriad of wavelengthsproduced and reflected by a subject end upbeing part of a digital image depends on anumber of factors. Before we explain theeffects produced by various types of IR filter,let’s take a look at the hurdles light enteringthe lens has to negotiate before it actuallyreaches the sensor.

‘Sieved’ Light

The glass of the lens elements cuts off a lot ofthe shortwave ultraviolet (UV) frequencies,

although the high-end coatings applied tomany lenses have a two-edged effect. Whilethey increase translucency by reducingreflection, they only work for a limited range ofvisible wavelengths. Coatings that work forother wavelengths are only available inexpensive special-use lenses.

Conventional coatings have very littleeffect on IR light. In some cases, they caneven create unwanted ‘hot spots’ in thecenter of the lens. This is because anti-reflexcoatings work by causing deliberateinterference and appear thinner to light

waves hitting them at an angle. This reducestheir effectiveness and can even amplifyreflections. The result is that more IR lightenters the center of the lens than the edges.Wide-angle lenses set to wide apertures areparticularly susceptible, but stopping downcan help to reduce hot spot effects.

The next obstacle for IR light is intentionaland takes the form of an IR cut-off filter (orsimply ‘IR filter’) located in front of the sensor.IR light hitting the sensor directly wouldcreate not only a significant red color cast butalso a blurred ‘ghost’ image.

114 c’t Digital Photography 18 (2015)

Infrared Photography | Filters

Infrared Filters and Their Effects

300 400 600 700 800 900 1000 1100 1200 nm500

300 400 600 700 800 900 1000 1100 1200500

Visible lightUV light Infrared

Solar Spectrum

Camera with IR cut-off filter

IR filter

IR filter

IR filter CMOS sensitvity

Blue Bayer filter

Red bayer filter

Green Bayer filter

Blue Bayer filter

Red bayer filter

720 nm IR pass filter 720nm IR pass filter

720nm IR pass filter

Green Bayer filter

Blue Bayer filter

Red bayer filter

Green Bayer filter

Camera without IR cut-off filter

300 400 600 700 800 900 1000 1100 1200500

300 400 600 700 800 900 1000 1100 1200500

300 400 600 700 800 900 1000 1100 1200500

300 400 600 700 800 900 1000 1100 1200500

300 400 600 700 800 900 1000 1100 1200500

IR filter Bayer filter

IR pass filter

CMOS

Bayer filter

IR pass filter

Red bayer filter

Green Bayer filter

Blue Bayer filter

Red bayer filter

Green Bayer filter

Blue Bayer filter

Each filter layer cuts off some of the light entering the lens. If an additional IR pass filter (in this case with a720nm threshold wavelength) cuts out invisible light too, virtually no light will reach the sensor in a conventionalunmodified camera (left). In a camera without a built-in IR filter, the sensitivity to IR light is determined by thelimits of the sensor’s own sensitivity (right). These diagrams are based on typical curves.

Page 8: Infrared Issue18 CtDP 2015

The shorter the wavelength of incident light,the ‘bluer’ it will be and the more it will berefracted on its way through the lens. Known as‘dispersion’, this effect causes the chromaticaberrations that lead to fringing effects,especially in wide-angle and/or poorlycorrected lenses. In order to capture a sharpimage in infrared light you therefore have toreduce the focus distance slightly. The distancescale on some manual-focus lenses evenincludes a separate IR distance mark to help youre-adjust focus once you have set up a shot.

Theoretically, you have to use a smallaperture to create a sharp image usingwavelengths that range from short (blue) toinfrared, but using too small an apertureincreases the risk of diffraction blur – and thegreater the wavelength of the incident light,the greater the risk. As you can see, it is almostimpossible to cater for all the incidentwavelengths, and an IR cut-off filter is aprudent tool to use to remedy some of the negative effects. To make everyday IRphotography flexible, practical andworthwhile, there is no real alternative but toremove the camera’s built-in IR filter (see page117), although doing so also causes a new setof issues that have to be addressed. Many IRphotography techniques are similar to thoseused by astrophotographers, who also haveto deal with light waves at the limits of thevisible spectrum.

In spite of all these limitations, it is stillpossible to capture IR photos using astandard, unmodified digital camera. Theattributes of built-in IR cut-off filters vary frommanufacturer to manufacturer, and most still

allow some NIR frequencies to pass. Mostvisible frequencies have to be cut out too, andthe best way to do that is by using anadditional screw-on or slip-on IR pass filterwith a specific threshold wavelength. Becausepass filters and cut-off filters work againsteach other, you need to use extremely longexposure times to compensate – the ‘better’the cut-off filter the longer the resultingexposure time will be.

The Bayer pattern microlens/microfilterarray makes up the final filter layer. If this wereoptically perfect (i.e., it really did only allowpure red, blue and green light to pass), cameramanufacturers wouldn’t need to use cut-offfilters at all, but then using conventionaldigital cameras for infrared photographywould be out of the question too. Fortunately,the infrared translucency of red Bayer patternmicrofilters drops off slowly and the greenand blue microfilters allow IR frequenciesthrough again at wavelengths above 700 nm.

The photoreceptors in the sensor are thefinal link in the filter chain. These are notsensitive to all light waves that reach them.Silicon-based photoreceptors can ‘see’wavelengths of up to about 1,400 nm (CMOSsensors are slightly less sensitive than CCDs),which makes them unsuitable for use inpurpose-built infrared cameras.

These are the limits within which we cannow begin taking infrared photos.

Which Filter Is Best?

At 750 nm, the Wood Effect is about aspronounced as it gets, and the reflections

caused by vegetation don’t get muchbrighter, even at longer wavelengths. If an IRpass filter has a threshold wavelength of750 nm, vegetation appears increasinglymonochrome and detail-free, producingfantastic, fairy-tale landscapes like the onereproduced on page 112. Be careful not toselect too high a threshold value – the effectof a 1000nm filter is not much different fromthat of a 850nm filter, but will reduce contrastand often sharpness too. The risk of producingunwanted hot spots also increases with thethreshold wavelength. In order to cut throughmost haze, wavelengths of 800 nm or less willusually suffice.

Captured using a filter with a threshold of720 nm or less, vegetation will still appearmonochrome but will be much more detailed.The more infrared ambient light in a scene,the weaker the filter needs to be in order toproduce a visible Wood Effect. 645 nm isenough in bright sunlight, while 750 nm is abetter choice in overcast weather. Becausethey require relatively short thresholdwavelengths, color IR photos are virtuallyimpossible to capture in the shade.

But not all filters are the same, even if theyare engraved with the same numbers. Theseonly indicate the wavelength at which a filter shows 50 percent of its maximumtranslucency but don’t tell us anything aboutits transmission curve. A 720nm filter mayallow more visible light through than a 715nmfilter. In absorptive filters, the steepness of thecurve depends largely on the thickness of theglass the filter is made of. The thinner theglass, the shallower the transition between

115c’t Digital Photography 18 (2015)

A ‘hot spot’ ensues when a lensallows perpendicularly incident IR

light through but blocks most of theIR rays entering at an angle

(i.e., at the edges of the lens)

Olympus E-PL-1 | M.ZUIKO 75mm f/1.8 |700nm conversion | ISO100 | f9.0 | 1/40s

Page 9: Infrared Issue18 CtDP 2015

the cut-off and translucency zones. If you arereplacing your camera’s built-in filter, a newfilter cannot be thicker than the one itreplaces. Interference filters achieve theireffects using coatings etched onto thesurface, so their effectiveness is independentof their thickness. Interference filters are,however, only usually used for specializedastronomical applications.

Long-pass filters are the ones most widelyused for IR photography. These block all lightbelow their threshold frequency and allow all longer wavelengths to pass through.Threshold values lie between 600 nm (deepred) and 1000 nm. Heliopan makes filterswith values of 610, 645, 665, 695, 715, 780,830, 850 and 1000 nm and prices start ataround US$130 for a 58mm model. Otherwell-known filter manufacturers offer justone or two models – for example, B+W (695and 830 nm), Hoya (720 and 900 nm) andCokin (720 nm). The clip-in filters fromastronomik.com (see page 119) come in 742 and 807nm versions. Astronomik hasannounced that ‘band pass’ filters speciallydesigned for IR phototography will beavailable soon. These block all light exceptthat within the 650-850nm waveband.

To use a camera that has been modified forIR use for conventional photography, youneed to use a cut-off filter. These are usuallyavailable as combined UV/IR filters from avariety of manufacturers. Astronomik sells

clip-in OWB (Original White Balance) filters for Canon APS-C and full-frame cameras that enable a camera to simulate its originalunmodified white balance behavior. Thesefilters start at around US$160.

Special-Use Filters

Some UV pass filters are transparent to IRwavelengths and therefore function as bandpass filters for two separate wavebands. Thisenables us to retain blue skies in color IRphotos without having to swap the red andblue channels (short wavelengths end uplargely in the blue channel anyway). TheHeliopan UG1 and Hoya U 360 filters aremade from Schott UG1 glass and are suitablefor this type of application. However, thereis probably no affordable lens available thatis capable of focusing all the requiredwavelengths on a single point and someparts of the frame will always end up out offocus. If this applies to the sky and theclouds only, it shouldn’t be too conspicuousin the final image.

Our experiments also revealed that somevariable neutral density (ND) filters can alsobe used as IR pass filters. We were originallylooking for a filter that prevents overexposurewhen shooting with the Sony DSC-F828 inNightshot mode, but neither of the filters wetested fitted the bill. They do block visible

light as planned but remained transparent toIR waves no matter how we positioned the two rotating polarizers. According to test measurements provided by themanufacturer, the Rodenstock Digital Pro MCVario ND filter has a threshold wavelength of810 nm in its fully ‘closed’ position. For the Quenox Fader model (available for about US$35 from enjoyyourcamera.com),our own tests revealed a thresholdwavelength of around 830 nm and a shallowtransmission curve.

The unique attribute of these filters is thatyou can add visible light to the IR light youcapture at will, simply by increasing thedegree of translucency. The downside of thisapproach is that it adds wavelengths from theentire visible spectrum to the mix, and the riskof short-wavelength blur once again rears itshead. The relatively high thresholdwavelength is also less than ideal forcapturing color IR photos. We solved bothissues by combining a variable ND filter witha 645nm pass filter, thus creating a hybridfilter that we could precisely fine-tune atwavelengths between 645 and 810 nm.Unfortunately, the effective range of thissomewhat clumsy setup is reduced to just afew degrees. Additionally, the Quenox Faderproduced inconsistent color errors in its‘closed’ position. Our wish for an effective IR pass filter with a variable thresholdwavelength remains unfulfilled for now.

116 c’t Digital Photography 18 (2015)

Infrared Photography | Filters

The curves on the graph show thedegree of translucency of the

Rodenstock Digital Pro MC Vario NDfilter when open (orange), half open

(green) and closed (blue)

MC Vario ND open (f6.3 | 1/500 s) MC Vario ND half open (f6.3 | 1/160 s) MC Vario ND closed (f6.3 | 1/125 s)

2000

10

20

30

40

50

60

70

80

300 400 500 600 700 800 900

Tran

sluce

ncy (

%)

Page 10: Infrared Issue18 CtDP 2015

117c’t Digital Photography 18 (2015)

Shooting | Infrared Photography

The amount of effort required to capture IRphotos depends on the type of equipmentyou have to hand. The quality of the results,your choice of filter and overall ease ofhandling depend very much on whether yourcamera still has its built-in IR cut-off filter, asthis will block most of the IR light entering thecamera (see the diagram on page 114). If youdo decide to remove the filter, there arevarious approaches you can take.

Amplifying Residual Light

Even with the IR cut-off filter intact, mostdigital cameras still let in a small amount ofresidual IR light at wavelengths above 700 nm.You can check whether this is the case for yourcamera by pointing an infrared remotecontrol at it and pressing a button. If the IRdiode on the front of the remote is visible onthe camera monitor, your camera is capableof capturing IR photos in its stockconfiguration. Begin your experiments usinga pass filter with a threshold wavelength of720 nm or more. If you use one with a shorterthreshold value, the visible light entering thecamera will blot out the IR waves.

You can then amplify the residual IR lightby using long exposure times and/or high ISOvalues. For our test subject, our EOS 60Dselected an exposure time of 1/40 s whenused with a 720 nm filter and set to f8 and ISO3200. This represents an increase of more thaneight f-stops compared with the exposureparameters the camera selected without thefilter (f8, 1/400 s, ISO 100).

Sony’s ‘Dark Angel’

Various manufacturers have sold a number ofcamera models without IR cut-off filters overthe years, although these were often hard tofind and weren’t available in all markets. Oneexception was the widely available SonyDSC-F828 bridge camera, introduced in 2003.In its special ‘Nightshot’ mode, the cameramoves the built-in IR cut-off filter out of thelight path. This unique mode was designedfor capturing photos in the dark and quicklyearned the camera the ‘Dark Angel’ moniker.Two IR LEDs built into the camera’s pop-upflash (or accessory IR lights) illuminate ascene, and the viewfinder and monitordisplay a low-resolution but adequatelybright live view image.

Used with a filter to block visible light, thiscamera appears ideal for daylight IRphotography. However, Nightshot mode isonly available in Full Auto and P exposure

modes, and the white balance and aperturesettings cannot be adjusted. The shortestavailable exposure time is 1/30 s, makingoverexposure inevitable in daylight situations.To counteract these shortcomings, you needto use an ND filter that is precisely suited tothe prevailing conditions and which doesn’tproduce over-long exposure times. You canalso experiment with a tripod or higher ISOvalues to compensate, although doing so willof course increase noise. Though no longer intune with today’s technology, the 8megapixels the camera’s sensor resolvesdeliver adequate resolution. Nightshot modeworks in RAW mode too, although it takes 15seconds to save each image! On the upside,the camera has a high-quality built-in Zeiss35-200mm f/2-2.8 lens that focuses veryquickly, even in pure IR light.

The camera is based on a 2/3“ CCD sensorwhich, instead of the usual three RGB colormicrofilters, has four ‘RGBE’ microfilters,whereby ‘E’ stands for ‘emerald’. This makesthe camera particularly sensitive to blue tones.In IR light, the sensitivity of the blue channelbegins at wavelengths below 700 nmn, so it iseven more sensitive to IR wavelengths thanthe red channel. This makes it difficult toseparate the various wavebands involved. Tocapture color IR photos, you need a relativelyweak IR filter that allows visible reds to pass –these tones then become blue when youswap channels during post-processing. Youcan also use a stronger IR filter to produceblack-and-white images with very highcontrast.

Infrared Camera Mods

The most effective – but also the mostradical – approach is to remove the cut-offfilter from your camera. This gives you an IRcamera that is easy to handle and guaranteesgreat results. If you already own a camera thatyou are prepared to cannibalize, having itconverted by a specialist shouldn’t cost morethan a few hundred dollars

You can, of course, dismantle your camerayourself. There are many camera moddingguides available on YouTube, although it is

questionable whether this approach is worththe effort and the risk. To be on the safe side,we recommend paying a specialist to do thework for you.

Simply removing the filter shifts the focalplane, so you also have to insert a clear glass sheet with exactly the same refractiveindex in its place or adjust the position of the sensor accordingly. As far as we know,only authorized Canon service centers are in a position to do this effectively.Phase-detection autofocus no longer worksproperly in IR light but can be adjusted.Alternatively, you can use live view, as thisuses contrast-detection autofocus. Thisapproach makes more sense if you are usingan IR filter mounted on the lens, as thismeans you can’t see anything through theviewfinder anyway. However, if you mountan IR filter directly in front of the sensor, you

720nm IR filterISO 3200 | f/8 | 1/40 s

NormalISO 100 | f/8 | 1/400 s

IR Photography with Conventional and Modified Cameras

Used with an IR filter and long exposuretimes, a conventional camera can capture

images of IR light waves. The original colorphotos were captured using a daylight

white balance setting and we increasedcontrast in the the black-and-white versions

to improve detail rendition.

Page 11: Infrared Issue18 CtDP 2015

can use the viewfinder as usual. As you cansee, the various approaches all have theirpros and cons. The following sections go intomore detail on the advantages anddisadvantages of each.

Neutral, Non-specificConversionThe IR cut-off filter is removed and replacedwith clear glass or the sensor is realigned (forCanon cameras only).

Advantages: The camera is sensitive tothe entire spectrum from UV to IR, andspecific wavebands can be selected usingappropriate filters. Varying Wood Effects canbe achieved using different grades of IRfilter. This type of converted camera can alsobe used to capture UV and astronomicalphotos and can be used conventionally withan additional IR filter.

Disadvantages: You need at least oneadditional filter, if not several to fit yourvarious lenses. With the exception of theAstronomik clip-in filters (see opposite) thatare available for a range of Canon camerasbut don’t work with all Canon lenses, thefilter has to be mounted on the lens.Because the filter is mounted in front of themirror, you cannot use the viewfinder andyou have to use live view instead.Phase-detection autofocus can only beadjusted to work for one waveband,whereas the contrast-detection autofocusused in live view mode works for allwavelengths, even if it is often painfullyslow. Mirrorless compact and system cameraare built to work in permanent live viewmode and don’t suffer from these issues.

Specific Conversion

An IR pass filter replaces the IR cut-off filter.Advantages: You don’t have to mount

additional filters on the lens and DSLRviewfinders work as normal. The fasterphase-detection autofocus functions too, buthas to be adjusted to work with the newconfiguration.

Disadvantages: The camera can only beused to capture IR photos and you are limitedto the threshold wavelength of the newbuilt-in filter. If you take this route, use a filterwith the lowest threshold wavelength you arelikely to need – you cannot weaken the effectany further later on, although you canincrease it by mounting an additional filter onthe lens. For example, a built-in 645nm filteris perfect for shooting color IR photos and allyou need is an appropriately graded IR filteron the lens if you want to capture high-contrast monochrome IR images.

118 c’t Digital Photography 18 (2015)

Infrared Photography | Shooting

Sony DSC-F828 | Nightshot mode/B+W 093 (830nm) | ISO200 | f2.2 | 1/30s | Black-and-white conversion and contrast increase performed in Lightroom

The legacy Sony DSC-F828’s built-in Nightshot mode makes it unusually IR-sensitive, especially in the blue channel

The ‘Nightshot’ mode built into theSony DSC-F828 saves you the troubleand expense of modifying yourcamera. Although limited in scope,this mode enables you to get out andstart taking your own IR photosimmediately.

Sony F828 in Nightshot mode

Sony F828 Nightshot + 695 nm Filter

Page 12: Infrared Issue18 CtDP 2015

Shooting | Infrared Photography

The thickness of the built-in filter is limitedby the space available within your particularcamera, which can lead to difficulties infinding a suitably high-quality filter (see page115 for more details on filter thickness andtransmission curves).

Real-world Tests

We were lucky enough to be able to test fivedifferent camera conversions, made by fourcompanies who specialize in this type of work.We compared these with our own convertedNikon D70 and the Sony DSC-F828.

As previously mentioned, convertedDSLRs suffer from clunky live view-basedhandling. In the Canon cameras we tested,autofocus was terribly slow and powerconsumption increased enormously. Thismade the camera heat up and produce moresensor noise. If you don’t want to use liveview, you have to set exposure manually and,if focus is set up for use in visible light, youhave to focus manually too.

These issues don’t apply for the OlympusMicro Four Thirds cameras we tested.

The clip-in filters from http://www.astronomik.com/en/ are available to fit a wide range ofAPS-C and full-frame Canon bodies but cannot be used with EF-S lenses, which protrude toofar back into the camera body

Page 13: Infrared Issue18 CtDP 2015

Although these cameras are not known fortheir super-fast AF, we were able to shootspontaneous snapshots in visible and IR light.The pale, monochrome IR monitor image canbe quite difficult to see in daylight, so theoptional electronic viewfinder is a usefulaccessory to have. In this respect, convertedDSLRs with an IR filter located directly in frontof the sensor are easier to use because youcan always see an unfiltered view of thesubject in the viewfinder.

The non-specifically converted EOScameras used with the Astronomik clip-infilters were easiest to handle and representone of the cheaper ways to shoot IR with aDSLR. These cameras don’t require additionallens-mounted filters, adapter rings, lens hoodsor lens caps and can be used conventionallytoo with an OWB filter. The only limitation hereis that you can only use EF lenses (EF-S lensesprotrude too far into the camera body andwould interfere with the clip-in filter). At thetime of writing, Astronomik wasn’t able tosupply a clip-in filter with a thresholdwavelength of 700 nm or less, which is why theOlympus PEN E-PL1 with its specificconversion ended up being our favorite testcamera. On the downside, two of our favoriteOlympus prime lenses showed obvious hot

spots when used with this model. These wereunacceptably pronounced with our M.ZUIKO75 mm f/1.8 and just about bearable with ourM.ZUIKO 12mm f/2.0. We produced muchbetter results using a Sigma 30mm f/2.8 andwe didn’t bother testing zoom lenses at all.

To retain the blue of the sky in IR photos, acamera’s RGB curves have to differ sufficientlyin both the visible and IR spectra. In thisrespect too, the Olympus cameras proved to be really useful, with a blue channel thatonly cuts in at wavelengths above 800nm andis therefore well suited to conversion to (infra-)red. The curves of the Canon EOS1200D/Rebel T5 and 600D/Rebel T3i are very similar but, even under near-identicalshooting conditions, we still found it difficultto create IR photos with punchy butnevertheless appealing colors.

The Canon EOS 60D loaned to us by anauthorized Canon dealer proved particularlytricky to use. Although the IR cut-off filter hadbeen removed, the sensor captured virtuallyno light at wavelengths above 700 nm. Othercompanies that specialize in IR conversionsalso confirmed the same issue with thisparticular camera model. We were unable tofind a satisfactory explanation for the problemby the time we went to press.

Our own measurements revealed atransmission curve similar to that in acamera converted for astrophotography.This type of conversion is not suitable for IRphotography because it extends thespectrum the camera can capture onlyslightly. The curve is much steeper at the redend of the spectrum, making it much easierto photograph galaxies that emanate redlight. Canon offers its own factory-madeastrophotography cameras such as the EOS60Da. (All such models are characterized byan additional ‘a’ in the name.)

With the exception of a couple ofcamcorder models, Sony’s ‘Nightshot’ featurehas disappeared from the market, which is ashame. The technology it utilizes would beperfect for capturing photos across a range of spectra if it weren’t for the technicallimitations built into the devices that supportit. We would also like to see Astronomik’sclip-in filters with a broader range of thresholdwavelengths and for other camera systems,such as Nikon, Sony or Micro Four Thirds.

It would be great if the IR market couldgrow enough to give filter makers and nichemanufacturers like Astronomik an incentive tocreate products that fill some of the gaps wediscovered.

120 c’t Digital Photography 18 (2015)

Infrared Photography | Shooting

Image Format

Always shoot in RAW if you can. This makesit possible to alter white balance aftershooting and makes it a lot easier to retrievehighlight and shadow detail that would belost forever in a JPG image. The ability toprocess images using 16-bit color depth isalso important when you are capturing andediting color IR photos.

ExposureJudging exposure is quite difficult basedon the strange colors that characterizeunprocessed IR preview images.Fortunately, auto exposure works fairlyprecisely in live view mode and you willusually only have to correct exposure by afraction of an f-stop in order to preventselective overexposure in a particular colorchannel. If you don’t use live view, theexposure parameters selected by thecamera can be completely off-kilter, so youare better off setting exposure manually.Always check the histogram, butremember that this is based on the camera

white balance setting. If you apply a newwhite balance setting later, you may findthat your exposure is too dark, which isbad for noise reduction but better thanoverexposure.

Because longer wavelengths are subject togreater refraction, you won’t need to stopdown as much as you would for aconventional exposure. In the case of IRphotography, f8 is not the universalsolution. If you are using a crop-formatcamera, the optimum aperture is actually alot smaller than the f-number would leadyou to think, so you can usually shoot wideopen – provided, of course, that thisprovides sufficient depth of field.

Exposure times in modified cameras arenot much longer than those required fornormal daylight photography. If you areworking with an unmodified camera andfilters, exposure times will be much longerand you will definitely need a tripod.

White BalanceEven if you shoot RAW, the camera whitebalance setting still affects the monitordisplay. A neutral- or blue-toned monitorimage is much easier to see than thereddish-brown one that results if you selecta daylight WB setting. The camera’sexposure meter, too, works in conjunctionwith the WB setting. If the exposureparameters give increased emphasis toone particular color channel (visible in thehistogram), the final image will end upunderexposed. Some histograms also havebuilt-in overexposure warnings, althoughthe risk of overexposure in an IR context isvery slight.

The best WB setting to use is the one thatprovides three similar RGB histogramcurves. The best way to achieve this is toset white balance using a sheet of whitepaper or sunlit grass as a reference. If youare shooting JPG, setting white balancemanually is essential, as you cannot alteryour images later without reducing theirquality.

IR Shooting Tips

Page 14: Infrared Issue18 CtDP 2015

If a scene doesn’t contain any objectsthat make a skewed white balancesetting too obvious, you can use oddballparameters to create highlyatmospheric color infrared photos

Olympus PEN E-PL1 | M.ZUIKO 12mm f/2.0 |700nm conversion | ISO200 | f7.1 | 1/200s

Page 15: Infrared Issue18 CtDP 2015

122 c’t Digital Photography 18 (2015)

Infrared Photography | Post-processing

If you are shooting black-and-white IR images,you won’t have to do much post-processing.Especially if you use a strong IR filter, the RAWimage captured by the camera will be largelymonochrome anyway. However, the whitebalance setting you use will strongly influencethe differentiation between the individualgray tones in your images and using thewrong setting can easily cause burned-outhighlights.

You should apply manual white balance as described on page 120 during RAWconversion at the latest. You can then convertthe largely colorless image to grayscale. Someimages work well if you select only thehighest-contrast channel and discard theothers completely.

If, on the other hand, you are aiming toproduce a color IR image with only partiallyshifted colors, the ‘channel swap’ techniquedescribed on page 112 is a better option. Themore selectively the Bayer pattern microfilterswork, the more effective this technique is. Thetutorial on the next page gives some tips onhow to enhance the effect yourself.

Divide and Rule

Increasing the differences between theindividual color channels also increases colorcontrast. Usually, the red, green and bluechannels contain the color data that the Bayermicrofilters of the same color permit to reachthe sensor. If you subtract the contents of thegreen and blue channels from the red channel(before swapping the red and blue channels),you will remove all the color data from the redchannel that would have appeared in the othertwo anyway. This makes it easier to selectivelyadjust the remaining red tones (i.e., visible redand those at the lower end of the IR spectrum).

If, as in the illustration above, the limits ofthe green and blue curves are far enoughapart, it can help to subtract the contents ofthe blue channel from the green channel too. The green channel then contains onlyshort-wave IR color data with wavelengthssimilar to those of the red tones. Whether allthis theory can be put into practice and theprecise weighting of the individual channelsrequired to get good results, depends on the

type of camera you use and the number of IRfrequencies contained in the captured scene.

Tutorials and Examples

Our tutorial demonstrates some basic IRimage processing steps and the ‘channelswap’ technique. To save disk space, wedeleted the layer mask present in all theadjustment layers, as this was not required forthe steps involved.

‘Channel swapping’ is the simplest way to adjust the colors in color IR photos,although other, more complicated mixturescan be used to achieve even better resultsthat require little additional post-processing.

The test image we used for the tutorial wascaptured using an Olympus PEN E-PL1 with its

standard IR cut-off filter replaced by a 700nmIR pass filter. The filter lets a little visible lightthrough and, along with the blue tonescaptured under the influence of thelonger-wave blue Bayer microfilters, producedideal source material for color IR photos. The12mm lens we used produced a mild hot spotthat is visible as a slight yellow shimmer justto the right of the tree.

We set white balance to 2850 K in theoriginal image, although just about anycolor temperature should work in this typeof shot. To create usable IR source images, itis best to shoot RAW and save the results to a large 16-bit color space such as ProPhoto RGB for processing. Theadjustments are then made non-destructivelyusing adjustment layers.

Once you have swapped the redand blue channels, it is easier toseparate the individual colors ifyou reduce the blue value

Our test image is included on thisissue’s free DVD, which also includes a

Photoshop Action that automaticallyperforms the steps detailed in the

tutorial on the next page

Post-processing IR Photos IR spectrum (Auto WB)

Channel swap R B

Chanel swap and Difference blend mode

Page 16: Infrared Issue18 CtDP 2015

123c’t Digital Photography 18 (2015)

Post-processing | Infrared Photography

Processing Color IR Photos

Before

2⎢Optimize the colors: The Image > AutoColor command is another usefulPhotoshop tool that you can use to removeunwanted color casts. In our example,applying the command gives the clouds areally clean white look. Once again, wemade our settings via the Enhance PerChannel Contrast dialog, this time selectingthe Find Dark & Light Colors and SnapNeutral Midtones options and setting theClip values to 0. This last prevents thehighlights and shadows from being clippedtwice. If you prefer bolder colors, or if youwant to adjust your colors manually (whichis easier with bold colors), you can alwayshide this layer.

3⎢Swap the red and blue channels: Create anew Channel Mixer adjustment layer,delete its layer mask and make the red andblue channel settings shown in theillustration. This swaps the contents of thechannels and often turns the sky blue whilegiving vegetation a yellow or red tint. Savethe Channel Mixer settings as a preset tosave time on your next attempt. You can, ofcourse, try out and save your own ChannelMixer ‘recipes’.

Before

After

1⎢Optimize contrast channel by channel:The first step is to ‘stretch’ the RGBchannels so that the darkest and lightestpixels are represented by true black andwhite. To do this, create a new Levelsadjustment layer and click the ‘Auto’button while pressing the Alt key. Thisopens the Auto Color Corrections Optionsdialog. Select the Enhance Per ChannelContrast option. This has the same effect as the Image > Auto Tone command andextends the RGB curves to fill the entirewidth of the histogram curve. In our sampleimage, this makes the clouds almostcompletely white and gives the tree a blue tint.

After

Page 17: Infrared Issue18 CtDP 2015

4⎢Set contrast and the gray point: We used aCurves adjustment layer to adjust brightnessand improve global contrast. The settingsshown here darken the sky a little and givethe image a feeling of greater depth. Weused the gray point eyedropper to removethe color cast from the clouds, although thisshifted the tone of the sky toward green.

5⎢Increase vibrance: We then used aVibrance layer to subtly increasesaturation. The Saturation slider affects allcolors, while the Vibrance slider primarilyaffects green and blue tones, leaving redsand yellows largely unchanged. Theindividual colors can be even moreselectively adjusted using aHue/Saturation adjustment layer,although this tool’s Saturation sliderworks on a much coarser basis.

6⎢Adjust the colors: We used aHue/Saturation layer primarily to shiftcolors. A Hue adjustment of +43 gave thesky the deep blue tone we were looking forwhile changing the color of the leaves toyellow. This is a typical IR look. If you wantto retain the pink hue in your image, youonly need to shift the color of the sky onceyou have selected your basic colors. To dothis, use the On-image Adjustment tool (the ‘hand’ icon) in the Layers panel, andthen use the eyedropper that appears toselect the color you wish to change.

124 c’t Digital Photography 18 (2015)

Infrared Photography | Post-processing

Processing Color IR Photos

c