Sulley's eye in UV, VIS and IR

[Stefano]

Sulley (James Patrick Sullivan) is the blue monster in Monsters & Co. I had a peluche of him when I was a child, which then became one of Sugar's many toys (our dog). It was very well-made, and resisted for months, until one week ago Sugar was able to tear his head open. He managed to tear apart one of his eyes clean, and I kept it. I imaged it in visible light, IR TriColour and UV BiColour.

 

Gear:

Camera: Full-spectrum Canon EOS M

Lens: Soligor 35 mm f/3.5

 

Filters:

IR: Hoya R72

UV: ZWB2 (2 mm) + Chinese BG39 (2 mm)

No filters in the visible light image

 

Channels:

IR:

Red: 940 nm

Green: 850 nm

Blue: 730 nm

 

UV:

Red: 365 nm

Cyan: 340 nm

 

The light sources were all LEDs.

 

Visible light. f/8, ISO 100, 1/2 s exposure

 

IR TriColour:

 

UV BiColour:

 

Notes:

- All images taken at f/8 and ISO 100, with varying exposures;

- I set the camera to auto exposure, as it is easier and faster to work this way. If I understand correctly, IrfanView white balances images by re-weighting the channels, and this is why it can't set a UV WB. It doesn't create non-existing channels (this reminds me of this conversation). This "feature" is actually very useful when doing TriColour images, as it compensates for the different exposures of the channels, without altering the colors;

- The UV BiColour image doesn't look very good for two reasons: the first one is that running my 340 nm LED at 100 mA produces very little light, and considering I used ISO 100 and f/8, that meant 20 minutes of exposure, and they weren't even enough. This resulted in a very dark image that needed to be lifted up by several stops. The second reason is that, for reasons unknown to me, the program I use to stack the images (ImageStacker) doesn't work properly if one of the images has been processed in Photo Ninja. I had to process the dark 340 nm .CR2 image using Windows Photo editor (it handles raws, surprisingly), and this may have impacted the quality a bit. Also, the LEDs weren't in the exact same spot, and the shadows on the paper tissue below show this.

- The three images (not the single channels, but the finished images) were not taken exactly in the same spot, and cannot be stacked. By changing the filters I couldn't keep the camera still, and it moved a bit.

 

I will try to improve my technique. I have to make the tripod sturdier, and when I will have better filters (that actually screw on the lens) and a nice collection of light sources as well as filters, I will be able to do more. For now, not too bad.

[Andy Perrin]

Very interesting. For the bicolor, you may wish to try the Lab color space. Use an 365nm for the L and then the 340nm for the b. Usually contrast needs to be reduced in the b channel beforehand.

[colinbm]

You didn't use the tissue to WB ?

[Stefano]

Very interesting. For the bicolor, you may wish to try the Lab color space. Use an average of the two images for the L and then the two individual images in the a and b. Usually contrast needs to be reduced in the a and b channels beforehand.

Thanks. How can I do it? Do I need to play with MATLAB?

[Stefano]

You didn't use the tissue to WB ?

I think I actually used the paper behind. The visible image was white balanced in-camera, but looked too yellow and so I fixed it. The IR image was yellowish too, the 730 nm image was exposed a bit less, and so I fixed it too. The UV image was bluish. In this case both papers have a similar reflectance curve, and can be used both. PTFE would have been better.

[Andy Perrin]

Thanks. How can I do it? Do I need to play with MATLAB?

 

No, I tried it in photoshop actually? I did it more or less as described, but I had to do some careful work with Levels and Photoshop's autotone to get things centered properly and avoid a strong cast to the image. Here's my results. It's still two unique pieces of information but now the 365nm is the lightness and the 340nm is controlling the color, which goes blue/yellow.

 

 

Sorry if I further degraded the quality, this is meant to show what's possible along these lines, a "proof of concept" only.

[Stefano]

Interesting, so the 365 nm image is the luminosity (L), and then the 340 nm image is yellow (bright) and blue (dark). I guess this gives pretty unusual images in general.

[Andrea B.]

Weird and fun results!