Tidy IR solution for mirrorless cameras

Kolari Vision have been busy. They have introduced a behind-the-lens clip-in system for filtering which allows you to swap filters directly in front of the sensor. This is especially useful for lenses which don't allow you to put a fiter in front of the lens, such as fish-eyes. Of course some lenses have special filter slots built-in but these are inherently restricted.

I had wondered whether a full-spectrum converted mirrorless camera would revert to conventional visible light work by putting an infrared-blocking filter (aka 'hot mirror') in front of the sensor/lens. After all, unlike an SLR, the exposure sensing is done from the main sensor, not from a separate system, so it will still function as normal if the 'hot mirror' has the same characteristic as the original.

It turns out that Kolari have come to the same conclusion. If you have a mirrorless camera converted for full-spectrum photography, such as a Nikon Z series, you can then add a suitable 'hot mirror' flter in front of the sensor simply by clicking it into the Kolari magnetic filter system. In that case, your camera is 'conventional' again.

One important thing to bear in mind is that if you buy a new camera and have it converted (by any facility, not just Kolari) you will probably invalidate the original warranty. So best to wait a bit.

So, two links to explore: Kolari's clip-in filters and hot mirror replacement (here as an external filter).

And while you're there, Kolari have a fascinating article about the dreaded infrared hotspots, which answers quite a few questions!

Dune movie scenes shot in Near Infrared

One of the big cinematic releases at the moment is the second part of Denis Villeneuve‘s interpretation of Frank Herbert's SF Epic Dune.

Cinematographer Greig Fraser decided to use near-infrared (NIR) imaging to show the weird environment of the planet occupied by the film's uber-villains, the Harkonnens. He had used the technique before, on Zero Dark Thirty in 2012 and Rogue One: A Star Wars Story in 2016.

While the Australian movie BRINDABELLAS | edge of light in 2016 (see below) had used RED cameras configured for monochrome IR, Dune used an ARI Alexa camera but the basic premise is the same. The usual infrared blocking filter was removed and replaced with a 'black' infrared-pass filter.

The idea with Dune was to show the unreal environment the Harkonnen's inhabited. The first film had only shown interiors but the second part required exterior shots. One significant result of this technique is the surreal look of the characters, since NIR penetrates a few millimetres into skin (and the characters are hairless) and there is the well-known look of people's eyes and the inherent high contrast.

[Photo: Dune: Part Two Infrared Copyright © 2022 Warner Bros. Pictures, Inc.]

I can point you to the following for much more information: Variety, NoFilmSchool, Kolari and ARRI Rental.

This provides an opportunity to review some of the history of near infrared in feature films.

Infrared film was a useful tool in the motion picture industry as far back as the 1920s. Kodak had produced their first infrared ciné film stock in 1925 and by 1937 it was also available from Agfa and DuPont. Agfa's was the first of what was described as the modern infrared film in that it was not a panchromatic emulsion pushed into infrared sensitivity. The new films were only sensitive to UV and blue and then to extreme red and infrared. This simplified the filter needed and a Wratten #29 (deep red) was the most common used. Sometimes infrared film was used in the making of travelling mattes (used to replace backgrounds in shots) but more often it was used in black and white movies to allow night-time scenes to be shot during the day, a technique now known as 'day-for-night'.

Not all the artefacts of infrared images were welcome however, and special makeup (usually lipstick) and set painting often had to be applied. Sometimes foliage was sprayed with green paint to hide the Wood effect and prevent shifts in tone. Paramount even painted an entire back-lot 'Brownstone Street' in special blue-grey paint called infra-red blue so that it would look the same on both infrared and panchromatic stock. The 1941 DuPont film was welcomed by cinematographers because of its lack of Wood effect and the three apparently competing emulsions had actually found slightly different and complementary niches in this specialised application.

By the 1960s the movie industry was moving from black and white to colour and infrared's abilities for day-for-night shooting were obsolete. But occasionally infrared filming was used for artistic effect.

In the early 1960s there was a curious collaboration between the Cuban and Russian film industries resulting in an extraordinary movie called Soy Cuba (I am Cuba). The director was Mikhail Kalatozov, famous most probably for The Cranes Are Flying in 1957, and the director of photography was Sergey Urusevsky. The film is a cinematic tour de force, featuring several long single-take sequences which almost defy attempts to work out just how they were done.

Much of Soy Cuba was shot using infrared film, with characteristic bright foliage and dark skies. The film stock was actually manufactured for use by the Soviet military, so it was quite a coup for the production to access some of it.

More recently, movie-maker Mike Figgis has been experimenting with low light and infrared photography using consumer video cameras with Sony's Night Shot facility. His 2001 film Hotel includes scenes done this way, to such an extent that the actors in the scenes could not actually see each other during filming.

The director of 2006 movie Wristcutters (A Love Story), Goran Dukic, had intended to use Kodak Ektachrome Infrared extensively to provide the look of the film's afterlife for suicides setting. Kodak provided unique super-16 format stock for this purpose, but after shooting some tests Dukic decided to use post-production techniques rather than infrared film. Some of the test sequences were shown on the film web site and on the published DVD. The production eventually sold off their unique stock for $300 per roll.

In 2015 film makers Glen Ryan and James van der Moezel of silver dory productions in Australia released a movie to exploit the monochrome infrared abilities of the RED digital cinema camera, called BRINDABELLAS | edge of light. It was described as “the World’s first near-infrared feature” and was shot in 4k resolution. I wrote it up in a blog post at the time, and the movie is still available on their web site.

Since most NIR-converted stills cameras can now shoot movies as well, the scope for infrared movies has expanded greatly over the past century.

Filtering a full-spectrum camera

My full-spectrum camera, a FujiFilm IS-Pro, has no built-in filter and shoots from near UV to near IR unimpeded. This also means that when I put a filter over the lens it also affects what I see through the viewfinder. With, say, a 720nm filter I am unable to see through the viewfinder so would have to use a tripod. Sadly the Fuji's live-view is pretty-well useless.

Many people replace the high-pass IR-blocking filter over the sensor with a low-pass filter such as a 720 or 820 nm to get around this problem ... at the small expense of loss of versatility.

In recent years I have used a deep blue filter, which I can see through to frame. Blue filters usually pass a lot of near IR. You can use a red filter of course, as you might have done with film, but I find the blue filter sometimes produces an interesting colour balance with minimal post-processing. Interestingly, the auto-focus works most of the time, which helps.

I recently bought a specially-designed filter for a type of colour infrared photography from the American company, Kolari Vision. It's called their IR Chrome Lens Filter, which I'll come back to in a moment. This nudged me into looking at results of a number of filters with the full-spectrum camera.

First, here is the camera output without any filter. This is basically a 'normal' colour image but with infrared contamination. You can click on the images to make them larger.

Next is a minus-blue (ie yellow) filter. This can be used to emulate the old infrared Ektachrome film ... see this blog page for more details.

Next comes the red (#25) filter.

Now the blue filter. Different black and white results can be achieved by either removing saturation or by selecting individual channels. (This also applies with other filters of course.) The green channel is useful because, with a Beyer filter camera the green channel has twice as many pixels as the red or blue. With this filter I find I need to under expose (according to the camera) by 3 or 4 stops.

This is Kolari's IR Chrome Lens Filter, which gives a good approximation of the old Ektachrome images. However, it is not exactly the same so is not as useful for foliage health analysis: but it's not a bad approximation.

I had achieved good results using neutral density filters in the past, with Sony's Night Shot, since the ND doesn't apply at IR wavelengths. So I bought a variable ND filter, which is basically two polarising filters together. You rotate one with respect to the other in order to reduce the amount of visible light going through. In this case once I had frames with minimal density I simply rotated the outer filter until I could only just see anything then fired the shutter. Autofocus worked and by trial and error found the exposure change: in this case under by 4 stops. There is a little (false) colour information left but this method works best for a monochrome result.

Finally, for comparison, here is a 720 nm filter result.

One thing this experiment also showed me was how bad the chromatic aberrations are around the edges in the lens I am using, which are quite noticeable with colour shots but usually vanish when reducing to monochrome.

For more on this subject, here is Kolari's page outlining the characteristics of their various filters.

The physics of near-infrared photography

Klaus Mangold (a photographer), Joseph A Shaw and Michael Vollmer (who are physicists) have just published a paper, The physics of near-infrared photography in the European Journal of Physics. This is the best technical paper on the subject that I've seen since Clark's book Photography by Infrared (which went out of print in 1984).

The European Journal of Physics has a policy of making papers freely available for 30 days from publication, although you will need to set up an online account to access it.

Amongst other things the paper tells us that red wine, Diet Coke and even espresso coffee are transparent to near-infrared wavelengths.

This is the URL: stacks.iop.org/EJP/34/S51

The citation is Eur. J. Phys. 34 (2013) S51–S71

DSLR infrared sensor response

Camera manufacturers tend to be coy about releasing the spectral response of their sensors: they're (logically) more concerned about how the camera as a whole performs than how the sensor might deal with infrared (or UV).

A peek behind that curtain has come courtesy of Christian Mauer of the University of Applied Sciences in Cologne. His thesis Measurement of the spectral response of digital cameras with a set of interference filters (January 2009) is interesting in its own right but if you work your way down to Appendix A page 81 (section A.1.7) you will find the spectral response of a Canon EOS 450D without its IR blocking filter. It shows the usual bump in blue response just beyond 650 nm so that the respective responses of the red, green and blue channels at 800 nm are around 20%, 30% and 40% of the peak green response (at 540 nm) respectively. All channels come together at 20% at 850 nm and the response is monochromatic beyond that.

These figures are actually showing the spectral response of the sensor plus its Beyer filtering, since the underlying sensor is monochromatic and should have a smooth response curve peaking at around 600 nm and slowly dying away to over 1000 nm. However, this explains the strange colours when you take infrared images through a 720 nm filter and why ones through an 850 nm or longer filter are colourless.

Filter experiments: blue

It's taken me a while to get around to doing some serious experimentation with my FujiFilm IS-Pro. This is basically a FujiFilm S5 with the infrared blocking filter removed and while, when released, it was ludicrously expensive the price came down to clear stock. As far as I know it was the only commercially available infrared-capable DSLR since the early Kodaks.

With an SLR the difficulty when photographing using infrared film was being able to see through the viewfinder. Of course you could use a tripod but this wasn't always an option. Most of the time I used a deep red (#25) filter, which gave a good Wood effect but could still be seen through, albeit dimly. The problem continues with a DSLR, especially if you don't have a live view mode.

I thought I'd try filtration alternatives with the DSLR - red, yellow (minus-blue) and blue - as alternatives to opaque infrared-pass filters such as the Hoya R72. The plan was to be able to see through the lens while making the most of the sensor resolution, which means always using the green channel as it has twice the pixels of the others in the Bayer matrix. Of course the Bayer processing mixes things up quite a bit and in an ideal world I'd write my own RAW decoder ... but perhaps that's for later.

I took the opportunity of a holiday in southern Spain ... with lots of sun and plenty of foliage ... to try out using a blue filter, a #47 colour-separation filter made by Tiffen. The light path is like this ...

... because the blue filter does pass a significant amount of near-infrared. I think all blue filters do but I can only vouch for the Tiffen.

Initial experimentation suggested that setting the camera's auto-exposure to -3 or -4 EV brought the histogram into the centre for all channels and the autofocus would work. I could also shoot at f8 and ISO 100 since there was plenty of 'light' available. The lens I used was a Nikkor 18-200 mm f/3.5-5.6 zoom which appears completely immune to the notorious hot-spot.

This is the result straight out of the camera (apart from some desaturation of the blue channel).

The colour palette isn't as broad as the one you get using an R72 (less greens) but I thought it quite pleasant. And what about the green channel on its own for that pure infrared look?

Pretty promising I think.