[ Beneath the Waves ]

Uses of Multispectral Photography

article by Ben Lincoln

 

At a very basic level, multispectral imaging is useful for the same reason as our ability to see the different colours of the visible spectrum: it increases our ability to determine at a distance the properties of what we see. Just as we can make an educated guess about the health of a common plant by whether its leaves are green or brown, or tell the difference between basalt and granite based on their relative brightness, other areas of the electromagnetic spectrum have their own information to share.

This aspect is taken to an extreme in the form of "hyperspectral" imaging and spectrometry. Spectrometry and spectrum analysis don't "take a picture" like a camera would, but provide a continuous graph of the electromagnetic emission from or absorption by a sample of material across a range of the electromagnetic spectrum. Here's an example from the Spitzer Space Telescope, an infrared spectrogram from a planet orbiting a distant star. You can think of this graph as being a "single-pixel" image of the subject. Hyperspectral imaging performs this same detailed measurement, but in the form of a traditional 2-dimensional image similar to what a conventional camera would generate. This flood of data obviously provides an enormous number of possibilities in terms of representation to a human eye.

A related (but distinct) application for imaging outside the human-visible spectrum is that it often allows the viewer to see what would otherwise be undetectable. X-rays are the most common example - they let doctors see bone structure without surgery. Photography of gamma rays can help determine at a safe distance whether or not an object is radioactive.

Each part of the electromagnetic spectrum has its own applications, but I will focus here on the ones which I actually have the equipment to work with.

Infrared

Although it is commonly referred to as a single spectral band ("the infrared"), there are three distinct divisions of this part of the electromagnetic spectrum. There is so much popular confusion about this that I've written a separate article specifically on one of the major differences (Thermal versus Near Infrared).

The section of the infrared closest to human-visible light is called the "near infrared" (also "short-wave infrared" or "SWIR" in some circles). This is the type of infrared that most night vision systems use[1], as well as many remote controls. This is the only infrared light that most cameras with traditional glass lenses can photograph.

The most striking aspect of the near infrared is that living plants reflect it extremely brightly, even when dry enough to appear brown to human eyes. Nature photos using near infrared light are instantly recognizable due to this factor.

Earth's atmosphere scatters near infrared light much less than visible light, which means that a near infrared photograph of a distant area will generally be more well-defined, and less likely to appear hazy. Clouds will appear with great clarity, and the sky will usually be very dark.

Thermal imaging systems make use of far infrared (referred to as "long-wave infrared" or "LWIR" in some fields) and sometimes mid infrared ("medium-wave infrared" or "MWIR") to determine the heat of an object without touching it. You may be familiar with this use if you've seen any of the Predator films. Beyond the obvious utility of being able to see the difference in temperature between objects, thermal imaging is very effective as a sort of "super night vision" system. Every object in the universe whose temperature is above absolute zero (which everyone other than scientists can interpret as "every object in the universe") emits thermal infrared. This means that unlike standard (near infrared) night vision devices, thermal cameras are effective even in the complete absence of light. Deep inside a cave - or an abandoned missile silo, etc. - standard night vision devices are useless without the near infrared equivalent of a flashlight. A thermal imager would still provide a clear view of such an area with no external illumination required.

All three regions of the infrared are very useful to astronomers, and CalTech has an excellent guide to this topic.

Ultraviolet

The most well-known aspect of ultraviolet photography is the patterns on flowers which I mentioned in A Detailed Introduction. If you haven't done so already, I cannot recommend visiting Bjørn Rørslett's site enough for examples of this (in addition to the rest of his photography, which is also stellar).

Image Enhancement and Virtual Photographic Filters

Beyond treating near infrared and ultraviolet-A light as additional primary colours, their proximity to the familiar three means that they can be used to enhance images or provide additional information in other ways.

Forensic photographers make use of near infrared and ultraviolet-A photography in part because blood and other evidence can show up in those images even when it is no longer detectable using conventional methods. In addition, bruises, tattoos, and veins are often much more visible in the near infrared.

Archaeologists have discovered that multispectral imaging can reveal text and images thought to be lost forever. Scrolls blackened by fire and ancient written works which were erased so their parchment could be re-used have both been recovered using near infrared and ultraviolet-A photography.

Kodak has developed a modified digital camera sensor which makes use of near infrared light to provide much better low-light performance. Similarly, both near infrared and ultraviolet-A can be used to enhance visible-light images in a way similar to some traditional photographic filters. I've written up some examples in the Virtual Filters article.

Forensic/Archaeology Example Images

I read a comment once that forensic work and archaeology were the same field, except that the time span which had elapsed between the event and the investigation was longer for archaeology. Here are a few examples to illustrate how multispectral photography can be useful for both types of work.

In these first two sets of images, I stained a black nylon "rash guard" shirt and a pair of BlackHawk(!) gloves with blood[2]. Both of them look mostly innocent in human-visible and ultraviolet-A light, although there a couple of small clots that can be seen if you look closely. However, the shirt gives up its secret competely in the near-infrared - there are two enormous stains on it.

Bloodstained Shirt
[ R-G-B ]
R-G-B
[ NIR ]
NIR
[ UVA ]
UVA
[ NIR-G-B [3C] ]
NIR-G-B [3C]
[ Difference: B-NIR [Grey] ]
Difference: B-NIR [Grey]
[ Ratio: UVA To NIR [Grey] ]
Ratio: UVA To NIR [Grey]
[ (Ratio: UVA To NIR)-(Ratio: UVA To G)-(Ratio: UVA To B) [3C] ]
(Ratio: UVA To NIR)-(Ratio: UVA To G)-(Ratio: UVA To B) [3C]
     

The bloodstains on this shirt are almost imperceptable except in the near infrared and variations based on it.

Date Shot: 2010-01-30
Camera Body: Nikon D70 (Modified)
Lens: AF-Nikkor 80mm
Filters: Standard Set
Date Processed: 2010-12-16
Version: 1.1

 

A friend of mine referred to my BlackHawk(!) gloves as "murderer gloves" once[3], and it turns out he wasn't far from the truth. Although there is a clot that can be seen near the knuckle of the index finger on the left-handed glove, the real story takes some advanced image processing to squeeze out of these gloves: a large section of the palm/finger area on both gloves is smeared with blood! The areas which appear red in the last (normalized difference) image are the ones that are bloody[4].

Bloodstained Gloves
[ R-G-B ]
R-G-B
[ NIR ]
NIR
[ UVA ]
UVA
[ NIR-R-UVA [3C] ]
NIR-R-UVA [3C]
[ (Difference: NIR-B)-(Difference: R-B)-(Difference: G-B) [3C] ]
(Difference: NIR-B)-(Difference: R-B)-(Difference: G-B) [3C]
[ (Difference: R-G)-(Difference: R-B)-(Difference: R-UVA) [3C] ]
(Difference: R-G)-(Difference: R-B)-(Difference: R-UVA) [3C]
[ (ND (Pos): R-G)-(ND (Pos): R-B)-(ND (Pos): R-UVA) [3C] ]
(ND (Pos): R-G)-(ND (Pos): R-B)-(ND (Pos): R-UVA) [3C]
     

The bloodstains on these gloves only become highly visible with complex image processing.

Date Shot: 2010-01-30
Camera Body: Nikon D70 (Modified)
Lens: AF-Nikkor 80mm
Filters: Standard Set
Date Processed: 2010-12-16
Version: 1.1

 

Multispectral imaging can also help reveal information from fire-damaged paper. Here are some variations on a "before" shot I did of a piece of office paper written on using various methods. Besides the regular pens, pencils, and so forth, two of the ones that are hard to read due to my handwriting are an "IR Invisible" and "UV UnVis" pen from LDP LLC.

Even in this shot, it's interesting to note in which variations the various text becomes possible to read.

Paper, Pre-Charring
[ R-G-B ]
R-G-B
[ NIR ]
NIR
[ UVA ]
UVA
[ All Bands (Skew)-All Bands (Kurt)-All Bands (ADev) [3C] ]
All Bands (Skew)-All Bands (Kurt)-All Bands (ADev) [3C]
[ (FC-3C-NIR-G-UVA) - DCS:  (YUV-Variance 0.500) ]
(FC-3C-NIR-G-UVA) - DCS: (YUV-Variance 0.500)
[ (FC-3C-NIR-G-UVA) - RDCS:  (YUV-25 Percent-Variance 0.300) ]
(FC-3C-NIR-G-UVA) - RDCS: (YUV-25 Percent-Variance 0.300)
       

Regular office paper with various kinds of writing on it. Note that several more complex variations have been included to reveal the locations of some of the less-obvious text.

Date Shot: 2011-01-01
Camera Body: Nikon D70 (Modified)
Lens: Nikon Micro-Nikkor 105mm f/4
Filters: Standard Set
Date Processed: 2011-01-01
Version: 1.0

 

To simulate fire damage, I charred the paper using a heat gun. As you can see, I wasn't careful enough in some places.

Paper, Post-Charring
[ R-G-B ]
R-G-B
[ NIR ]
NIR
[ UVA ]
UVA
[ Kurtosis-Average Deviation-Minimum [3C] ]
Kurtosis-Average Deviation-Minimum [3C]
[ Skewness-Average Deviation-Median [3C] ]
Skewness-Average Deviation-Median [3C]

 

Date Shot: 2011-01-01
Camera Body: Nikon D70 (Modified)
Lens: Nikon Micro-Nikkor 105mm f/4
Filters: Standard Set
Date Processed: 2011-01-01
Version: 1.0

 

Note that while note all of the writing is visible at all in the near infrared shot, the writing that is there is much easier to read, because the paper is not nearly so darkened in that part of the spectrum. In addition, the damage is far more apparent in the ultraviolet-A shot. This may provide clues indicating that a piece of paper was exposed to heat even if its appearance has not changed to our eyes.

Multispectral photography is not a guaranteed method of capturing a better image. These petroglyphs, from Gingko Petrified Forest State Park in Eastern Washington, show up best in the regular version of the photo:

Petroglyphs
[ R-G-B ]
R-G-B
[ NIR ]
NIR
[ UVA ]
UVA
[ NIR-G-UVA [3C] ]
NIR-G-UVA [3C]
 

 

Date Shot: 2008-08-26
Camera Body: Nikon D70 (Modified)
Lens: Nikon Series E 28mm (?)
Filters: Standard Set
Date Processed: 2010-10-09
Version: 1.2

 

Can't You Just Do That in Photoshop®?

This is a fairly persistent belief which I'll attempt to dispel here.

Of course, any image can be "faked" digitally given enough time and persistence - and in fact, working with image-editing software is a big part of the process I use. The difference is that a simulated image won't actually reveal anything new about its real-world subject.

Photoshop®'s channel mixer includes a preset which converts a regular image into a greyscale simulated near infrared picture, and Ken Rockwell gives a guide for an alternate method. In all fairness, Mr. Rockwell makes clear that the instructions he gives are for artistic purposes, not an attempt to make a true equivalent of an infrared photo.

Here's a comparison between the real near infrared image and these two "faux" approaches using one of the images from my photos of Yellowstone National Park.

Faux NIR Greyscale
[ Real NIR ]
Real NIR
[ Faux NIR (Photoshop Preset) ]
Faux NIR (Photoshop Preset)
[ Faux NIR (Ken Rockwell's Method) ]
Faux NIR (Ken Rockwell's Method)
   

A comparison between one real and two simulated near infrared images.

Date Shot: 2009-09-01
Camera Body: Nikon D70 (Modified)
Lens: Nikon Series E 50mm (?)
Filters: Standard Set plus B&W 106 and 110 NDF
Date Processed: 2009-10-11
Version: 1.0

 

In greyscale, they look at least vaguely similar, although Ken Rockwell's instructions give a much more convincing result. However, when used in a false colour composite, the flaws become glaringly obvious:

Faux NIR False Colour
[ Faux NIR-R-G (Photoshop Preset) ]
Faux NIR-R-G (Photoshop Preset)
[ Faux NIR-R-G (Ken Rockwell's Method) ]
Faux NIR-R-G (Ken Rockwell's Method)
[ Real NIR-R-G ]
Real NIR-R-G
   

A comparison between one real and two simulated NIR-R-G false colour images.

Date Shot: 2009-09-01
Camera Body: Nikon D70 (Modified)
Lens: Nikon Series E 50mm (?)
Filters: Standard Set plus B&W 106 and 110 NDF
Date Processed: 2009-10-11
Version: 1.0

 

While the Ken Rockwell technique is much closer to the genuine article than the Photoshop® preset, it is of course unable to reveal the true near infrared signature of this shot. The microorganisms whose growth appears as red in the "true false colour" version are nowhere to be seen, because that distinction is completely invisible in the human-visible spectrum. The mud and water retain the brown appearance they have for human vision, and the plants which are brown instead of green in the original image (for example, near the center) lack the bright reflectivity they should display in the "faux" near infrared.

My Interests

I enjoy multispectral photography because it lets me see the world in a way that my eyes alone cannot. As long as I can remember, some of my favourite stories have involved worlds that are "just a few steps away" from our own. Charles de Lint's Jack of Kinrowan novels, the Soul Reaver videogames, and the Nightwatch series of books and films from Russia would be the most prominent examples. Reading about astronomy and other branches of science made me realize that there are truly many different views of the universe beyond the one that we tend to take for granted, and seeing LeVar Burton as Geordi LaForge use his "VISOR" prop on Star Trek: The Next Generation as a child got me thinking that these ways of seeing the world are actually within our reach thanks to technological advancements.

People are often at their best when they're exploring uncharted terrain, and exotic imaging techniques turn everything into a potential opportunity to step into a valley that's never been trod on by human feet, or turn over a stone that has never been touched by our hands.

 
Footnotes
1. Although modern night vision systems use image amplification and do not require an infrared light source except in very low-light situations, the technology has its roots in World War II-era devices (such as Nazi Germany's "Vampyr" sniper rifle scope) which were ineffective without a bright infrared illuminator. These original devices were less "night vision" and more "invisible light video camera", in that the advantage they conveyed was the ability of the user to shine a spotlight onto their target which the target would be unaware of unless they possessed a similar near infrared scope. Low-end consumer devices today still require the use of such an accessory, and even high-end night vision equipment will generally include one for use when there is not enough ambient light even with amplification.
2. A nurse at my doctor's office was kind enough to draw a sample of my blood for these purposes. Did you know that blood stored outside of the human body without a clotting agent clots so strongly that it looks like a cherry-flavoured carrageenan/gelatin dessert (which I am not using the trademarked name of to avoid getting a nasty letter from the manufacturer's PR/legal department)? Imagine that scene in Blade II in which the vampires in the nightclub in were eating jiggly, translucent red lumps with forks! Mmm! In case you are wondering, I asked for the anti-clotting agent to be left out because obviously blood left at a crime scene would not have such a chemical mixed with it, and I wanted the test images to be as authentic as possible.
3. Good thing I'm not actually a murderer, or maybe that one time would have been the only chance he got.
4. Much to my dismay, the fact that it's possible to make the bloody areas show up with such clarity in this particular image appears to be a quirk of that particular glove material. The same technique applied to the shirt in the first example does not have the same effect (almost certainly because the near infrared reflectivity is different), so there is no magic "BloodCam", at least for now.
 
[ Page Icon ]