• Ultraviolet Photography
  •  

My attempt at tri-color IR

Infrared Processing TriColour
24 replies to this topic

#1 Stefano

    Member

  • Members(+)
  • 1,800 posts
  • Location: Italy

Posted 26 November 2020 - 19:52

Inspired by Bernard's excellent work, I wanted to try full color/tri-color IR too. I didn't use three separate filters, but three separate light sources. More below.

I have a wide range of IR LEDs, currently seven between 730 and 1050 nm, but only three of them are usable as of now, since I didn't attach the others to a heatsink yet, and thus they would overheat. Those three LEDs are the most common IR LEDs you can find online, emitting at 730, 850 and 940 nm. Their peaks are roughly evenly spaced, and thus they are suitable for tri-color photography (one can use any combination he/she likes, but evenly spaced filters/light sources are better in my opinion). The LEDs are the "10 W" type, with 9 chips in a 3S3P configuration and a maximum rated forward current of 900 mA. They probably emit 1-3 W of light, not more. I may one day write a topic about my LEDs in detail. I ran all of them at full power.

The target was water. People who read my posts for a while know that I like seeing the absorption of water in the near-infrared spectrum (possibly, one day, even in SWIR), and since I know that water appears noticeably darker at 940 nm by experience I wanted to combine three images to make it appear blue.

I used my full-spectrum Panasonic DMC-F3, an Hoya R72 filter to prevent any possible (but unlikely) contamination by visible light, and to prevent movements between the images I mounted the camera on a tripod which I attached to the floor with bi-adhesive tape. Since I didn't care about colors in the single images, and I would have needed to convert them to B&W anyway, I directly shot them in B&W in-camera. To have uniform exposures between the images I put the camera on auto ISO mode, and it worked very well.

I put a paper tissue in the background to have a white target. Normal copy paper would have worked as well. The water thickness was 28 mm, and the LEDs were ~50 cm (~20 in) from the container. I mapped 730 nm as blue, 850 nm as green and 940 nm as red.

Images settings were f/2.8 for all images, and
1/30 s ISO 80 for 730 nm;
1/30 s ISO 100 for 850 nm;
1/8 s ISO 320 for 940 nm.

Combined final image (just the three channels stacked, no white balancing, no alignment, no post-processing):
Attached Image: somma 263.jpg

Increased saturation:
Attached Image: somma 263 saturazione.jpg

Any suggestion is welcome.

Edited by Stefano, 26 March 2021 - 22:06.


#2 Bernard Foot

    Bernard Foot

  • Members+G
  • 718 posts
  • Location: UK

Posted 26 November 2020 - 22:28

Stefano - good thinking: a totally different approach to the objective. I'm surprised how short your exposures were - those must be bright LEDs. You could easily photograph room-sized objects.

You certainly got the colour of the water right, and the white balancing looks about right. It might be worth increasing contrast - to get the background whiter and the floor blacker. That would probably bring the colour of the water out better.

It would be interesting to see your technique on other subjects that have more colour in them.

Are there suitable UV LEDs that could be used for UV tri-colour?
Bernard Foot

#3 Stefano

    Member

  • Members(+)
  • 1,800 posts
  • Location: Italy

Posted 26 November 2020 - 22:44

Thanks Bernard. The LEDs are bright enough to light up a room. At 1-2 cm, blue and shorter wavelengths heat up my skin. Not to the level of my UV LED, but you can definitely feel heat from radiation.

I may try to edit my image and post the results.

I will surely try this technique again with other subjects.

View PostBernard Foot, on 26 November 2020 - 22:28, said:

Are there suitable UV LEDs that could be used for UV tri-colour?
...yes, kind of. Power is the main issue. You can buy a set of three LEDs emitting at 310, 340 and 365 nm for example, but the 310 and 340 nm ones will be, as you may expect, weak. Also you want your light sources to have the same shape and size to project shadows in the same way. Using a single light source of course eliminates this problem, but this is not what I am doing. Two solutions could be to put the LEDs far away so that they will all be point sources of light or to use a diffuser to diffuse their light the same way. I have a 365, 385 and 405 nm LED, I may try those in the future (the 385 nm one is still not on a heatsink).

#4 Stefano

    Member

  • Members(+)
  • 1,800 posts
  • Location: Italy

Posted 26 November 2020 - 22:50

Also, there is another possible issue with UV LEDs: fluorescence. This of course doesn't occur in IR, where you have to look hard to find something emitting light from IR-induced IR fluorescence, but in UV this is much more common, so you would need at least a shortpass filter.

#5 Stefano

    Member

  • Members(+)
  • 1,800 posts
  • Location: Italy

Posted 27 November 2020 - 00:17

Ok. I played with the image a bit, and got this:
Attached Image: somma 263 corretta.jpg

I also tried with a Rubik's cube. This time I had to white balance, as you will see. The reason is probably because the cube was significantly darker at 730 nm, and the camera increased the exposure.

Visible reference:
Attached Image: P1000633.JPG

Tri-color IR:
Attached Image: somma 264.jpg

White balance on the paper tissue:
Attached Image: somma 264 WB.jpg

Next time I will have to find a cleaner place to take the images. The floor is not the best choice, but these are just initial tests.

#6 colinbm

    Member

  • Members+G
  • 2,418 posts
  • Location: Australia

Posted 27 November 2020 - 01:08

Wow, fantastic Stefano, excellent tests & results.

#7 Stefano

    Member

  • Members(+)
  • 1,800 posts
  • Location: Italy

Posted 27 November 2020 - 01:23

Thanks Colin!

#8 colinbm

    Member

  • Members+G
  • 2,418 posts
  • Location: Australia

Posted 27 November 2020 - 01:26

I can appreciate why you would select the three LEDs to be evenly spaced.
But the visible colours that we label, RGB & CY as in RYGCB are not evenly spaced & I have wondered why & it seems inconvenient ?
I intentionally left out M as it is not a colour of the rainbow, & colour is not circular as in a colour wheel.

#9 Andy Perrin

    Member

  • Members+G
  • 4,093 posts
  • Location: United States

Posted 27 November 2020 - 01:49

Nice work. The general approach of varying the light source rather than the filter to do tricolor images is something that i started working on for SWIR tricolor images. The advantage is that you can switch the LEDs on and off electronically without disturbing your photo subject between shots. So far, I have not been able to find the SWIR LEDs I want for under $1000 (cumulatively) with adequate protection for max current.

Colin, you are correct, the spectral colors in visible light are not evenly spaced in wavelength. The three cones that do the sensing, labeled L (long wavelength), M (medium), and S (short) do not have sensitivities that are evenly spaced. In particular the L and M are much closer to each other. (In my eyes, they overlap almost completely leading to the red/green colorblindness.)

Edited by Andy Perrin, 27 November 2020 - 01:54.


#10 dabateman

    Da Bateman

  • Members+G
  • 2,771 posts
  • Location: Maryland

Posted 27 November 2020 - 04:07

This site has some good information and the spectrum of an average eye:
https://askabiologis.../rods-and-cones

The rods don't see color, normally. But I have heard of individuals whom do see with the rods. Leads to odd perception.

Pedro was working on using UV leds. You can read some of it here:
https://www.photo-sp...otebook.nb.html

The older posts seem to have moved around as he switched blog sites.


#11 colinbm

    Member

  • Members+G
  • 2,418 posts
  • Location: Australia

Posted 27 November 2020 - 09:55

Thanks dave, very good articles to read.

#12 Bernard Foot

    Bernard Foot

  • Members+G
  • 718 posts
  • Location: UK

Posted 27 November 2020 - 11:17

Interesting shot of the Rubik cube, Stefano. Nicely done.

The colouring of yellow and brown corresponds to what I got when photographing plastics using the filter method.
Bernard Foot

#13 Stefano

    Member

  • Members(+)
  • 1,800 posts
  • Location: Italy

Posted 27 November 2020 - 12:28

Thanks Bernard.

Any suggestion for colorful IR objects? Blue is a rare color, especially in tri-color IR.

#14 Stefano

    Member

  • Members(+)
  • 1,800 posts
  • Location: Italy

Posted 27 November 2020 - 13:45

View PostAndy Perrin, on 27 November 2020 - 01:49, said:

[...]with adequate protection for max current.
Andy, if you don't have one, you should buy a bench CC CV power supply. With its current-limiting feature, you can drive any LED you like. I bought mine for around 50 €. You can even short them, and the current will still be limited to the value you set. Mine can do 0-32 V and 0-5 A, so it can deliver up to 160 W of power, which is plenty for most things.

#15 Stefano

    Member

  • Members(+)
  • 1,800 posts
  • Location: Italy

Posted 27 November 2020 - 13:50

Meanwhile... who wants to drink some blue water?

Note: I forgot to say I forgot to put the Hoya R72 filter on the lens, but I was in total darkness with blinds/shutters down and there shouldn't be any contamination.

Visible reference (sorry for the bad focus):
Attached Image: P1000635.JPG

Tri-color IR (no adjustments):
Attached Image: somma 265.jpg

Increased saturation + white balance (the white balance was already quite good anyway):
Attached Image: somma 265 c - Copia.jpg

I need some greens... I will look up Bernard's images to find some.

I wrote "Tri-color UV"... I meant IR of course! Just fixed.

Edited by Stefano, 27 November 2020 - 17:06.


#16 Bernard Foot

    Bernard Foot

  • Members+G
  • 718 posts
  • Location: UK

Posted 27 November 2020 - 13:57

View PostStefano, on 27 November 2020 - 12:28, said:


Any suggestion for colorful IR objects? Blue is a rare color, especially in tri-color IR.


Not really - I didn't find much colour in stuff you could photograph indoors. Water, as you know, comes out blueish, as does iso-propyl alcohol. Flowers, fruit, and vegetables come out almost white, with only light colouring. The only real colours I got were yellow and brown, generally in man-made things. I got a defintie red-brown from blue-glazed pottery.

You might try photographing people. My wife's skin came out pale blue, and her brown hair came out more reddish - just like it had been when she was younger.
Bernard Foot

#17 Stefano

    Member

  • Members(+)
  • 1,800 posts
  • Location: Italy

Posted 27 November 2020 - 14:07

I too wanted to photograph my hand, but it is impossible without messing things up... I will need someone else's hand.

In general, even natural things (such as flowers and fruits) are more colorful in VIS and even in UV than in IR, IR light isn't used by anyone and I think not even animals can see it, so most colors are "turned off" past red.

...if only some company manufactured a true tri-color IR sensor! That would be beautiful, but the market would be tiny.

Edited by Stefano, 27 November 2020 - 14:12.


#18 Stefano

    Member

  • Members(+)
  • 1,800 posts
  • Location: Italy

Posted 27 November 2020 - 15:34

For SWIR I would use three nice wavelengths: 1200, 1450 and 1650 nm.

...but Andy's TriWave doesn't see past ~1600, so I would use 1200, 1300 and 1450, or 1300, 1450 and 1550, or other combinations.

Thorlabs has some nice SWIR LEDs, which are similar to UVB and UVC LEDs in terms of efficiency, power and price.

1200 nm: https://www.thorlabs...tnumber=M1200D2
1300 nm: https://www.thorlabs...tnumber=M1300D2
1450 nm: https://www.thorlabs...tnumber=M1450D2
1550 nm: https://www.thorlabs...tnumber=M1550D2
1650 nm: https://www.thorlabs...tnumber=M1650D2

Weak? Yes. Under $1000? Also. They are little funny semiconductor squares.

#19 Andy Perrin

    Member

  • Members+G
  • 4,093 posts
  • Location: United States

Posted 27 November 2020 - 16:18

Stefano, I already know Thorlabs whole LED collection by heart. Trust me, if I thought those were adequately I would have them. The $1000 is the CUMULATIVE cost also, for ALL the parts required. Not individual LEDs. Please remember they need heat sinks and power and the ability to be switched on and off by computer. They also need protection against too much current being applied.

Edited by Andy Perrin, 27 November 2020 - 16:21.


#20 Stefano

    Member

  • Members(+)
  • 1,800 posts
  • Location: Italy

Posted 27 November 2020 - 17:04

I still think you can stay below $1000. These LEDs cost ~124 € each (around $150 if the price is the *same* for you, without extra costs), except for the last one which is more expensive (but your camera can barely see it). Three LEDs of your choice should cost about $450, let's even say $500. The heatsinks are not expensive, at most $20/30 cumulative. Thermal paste, screws and so on are cheap too. For around $50/60 you have a current limited power supply. You still can damage LEDs if you briefly short the terminals (there is probably an inductor in series with the output that gives you a voltage spike), or if you connect them to the power supply with a too high voltage, in both cases even with the current limited to a safe value. You can prevent short circuits, but you have to be more careful about the voltage the power supply has when you connect the LED, if it is just right or slightly above it is Ok.

Why do you need to switch them on and off by computer?