Jump to content
UltravioletPhotography

Infrared 3-band color images (780, 830, 1064 nm)


Andy Perrin

Recommended Posts

Not the first time this has been tried, but I attempted to make some wavelength-dependent false colors by taking three photos using a 780BP30, an 830PB40, and a 1064BP25 filter and putting the resulting images in the blue, green, and red channels respectively. The camera used was the TriWave, which is monochrome and has no Bayer filter to distort the results. Lighting was a halogen light with some kind of shield over it.

 

The physical setup looked like this. I have the TriWave attached to an iris, followed by a 100mm lens with a NIR/SWIR AR coating from Thorlabs, and then a sliding filter holder that lets me easily swap filters without messing up the image.

post-94-0-35768700-1579054705.jpg

 

The experimental subject was this jalapeño:

post-94-0-72859900-1579055094.jpg

 

The filter spectra (supplied by Omega with the filters) were:

post-94-0-10461900-1579057524.png

 

The three unaltered images came out like this:

 

Gain of the camera went down over the range, so I adjusted exposure time by roughly one stop for each image.

 

780BP30 (blue channel)

post-94-0-19176900-1579055254.jpg

 

830PB40 (green channel)

post-94-0-71307400-1579055266.jpg

 

1064BP25 (red channel)

post-94-0-42289400-1579055280.jpg

 

I took one additional image at 1500nm long pass (the end of the camera's range). This wasn't used for anything, I was just curious and it was the last filter in the filter holder, so I took it "while I was there anyway."

post-94-0-91785500-1579055384.jpg

 

I put the images in the three channels and got this:

post-94-0-99156700-1579055523.jpg

 

Then I whitebalanced off the paper in the background using PhotoNinja and trimmed the histogram for better contrast:

post-94-0-56028000-1579055678.jpg

Link to comment

Very cool Andy.

It looks kind of how I expected. Not too much difference in that range.

For you what maybe a better set to see some differences would be your 980bp10 filter, the 1064 filter and then the Lp1500. That should pull out color differently as the 1064 seems kind of equal to the 780 and 830 filters.

Link to comment

Wow, we have our first full-colour IR shot - although there's not much colour, as Dave predicted. (Actually, I've found the same in UV - there's not a lot of colour to be seen there either: pollinators only see UV at the longer UVA wavelengths, so no evolutionary benefit to develop colour at shorter wavelengths.)

 

Chlorophyll reflectance tails off only slightly from 700 to 1300nm so I wouldn't expect vegetation to show much colour - perhaps a tinge of blue or cyan, which might be suported by Andy's image.

 

But this is just the first image we have. Plenty more subjects to try out. I'm interested to see if there's any colour in the sky (i.e. any difference in scattering at various IR wavelengths) - perhaps a blue tinge. Manmade coloured objects could be interesting. Perhaps people (if you can get them to stay still long enough). And when the weather is obliging, I'm interested to see if snow has any UV and IR colour, or whether that is white across the whole range.

 

My initial filters should be here soon, so I'll be able to start. They will cover a different range to Andy's - blue channel will start at 700nm rather than 750, green will extend to about 900nm rather than 850nm, and red will start at 920 or 950-ish, depending on which filter I use.

 

I don't know how we could arrange it, but it would be interesting if we could photograph the same subject with our varioua filter sets and see if there is any difference.

 

 

It may be an idea to start a new thread for IR 3-band Colo(u)r Images that isn't specifically about 780, 830, 1064 nm, and put our examples and technical observations in there.

Link to comment

Dave - re. your ocmment on UG1 in the previous thread about ZWB1, this does look good in the 700-800 range, but it leaks a lot more than UG11 & ZWB1 in the 800+ range, so the 700-800 image would have a lot of 800+ contamination.

 

Graph shows UG1 (green), UG11 (red), China ZWB1 (blue) when combined with R72, taking account of decreasing CMOS sensitivity as wavelength increases. Vertical scale goes up to 30%.

 

post-245-0-05484500-1579091129.jpg

Link to comment

Dave - re. your ocmment on UG1 in the previous thread about ZWB1, this does look good in the 700-800 range, but it leaks a lot more than UG11 & ZWB1 in the 800+ range, so the 700-800 image would have a lot of 800+ contamination.

 

Graph shows UG1 (green), UG11 (red), China ZWB1 (blue) when combined with R72, taking account of decreasing CMOS sensitivity as wavelength increases. Vertical scale goes up to 30%.

 

post-245-0-05484500-1579091129.jpg

 

Yes but that peak is 20% for Ug1 and the tail is less than 1%. Your peak for U340 is what 5%?

 

If you overlay that graph with a typical sensor QE curve you will see it doesn't matter.

 

Like thus user calculated one for the sensor in my EM1 (Panasonic MN34230)

post-188-0-65547100-1579091695.png

Link to comment
The visible spectrum is the most beautiful. Lots of colors, even from nature. In the NIR everything basically bleaches out. The opposite happens in UV, everything absorbs more or less strongly. Of course there are exceptions, but that is the "trend" for most things. It isn't a coincidence that the sun emits mostly visible light, and our eyes evolved to see light which is more abundant on Earth. Wander what aliens see on planets around red giants, probably from green to NIR.
Link to comment

Dave - I think your post and my edit to add the graph to my previous post co-incided.

 

I've discarded the idea of the U-340. The options I'm most interested in are UG-11 and ZWB-1. Compared with these, UG-1 is good for transmission in 700-800, but poor in leakage at 800+.

 

Sterfano - I agree with you that there is far more and more beautiful colour in the visible range. But that's not the point of the exercise, of course - although it does help to explain why I am not prepared to throw too much money at this!

Link to comment

Do you want to see some color in IR?

 

https://www.google.com/url?sa=t&source=web&rct=j&url=https://www.spiedigitallibrary.org/journals/optical-engineering/volume-52/issue-04/045008/Study-on-shortwave-infrared-long-distance-imaging-performance-based-on/10.1117/1.OE.52.4.045008.full&ved=2ahUKEwj5w4jg2YXnAhWRzKQKHaliBlIQFjAEegQIBBAB&usg=AOvVaw0k7fM2uUhZduci7NnL8WNI&cshid=1579094656127

 

Although this includes SWIR up to 2140 nm.

 

And yes, Rayleigh scattering works in UV and IR. Someone posted something on the UV side, I may find the link. In the study I linked you can see blue sky in IR.

 

I hope the link works.

 

Figures 7, 9 and 12 are excellent.

Come on Andy lets see some distance images of Massachusetts.

 

Link to comment

Thanks for the link, Stefano - I've had just a quick look at present. So we may be seeing some colour ...

 

I notice though that the article uses far longer wavelengths than I am planning or able to use.

 

If Andy doesn't pull his finger out, I'm going to beat him to it (but Buckinghamshire rather than Massachusetts) ...

Link to comment
The one outdoor IR trichrome I have seen rendered open sky as a deep, midnight blue, almost black. That was using a filter at around 850 for the blue channel. YMMV, of course. Almost everything else was palest pastels or colorless.
Link to comment

Oh, I am going to do SWIR color over the range 1200nm-1500nm. That is what the above rig is really intended for, but I have to custom make my filters from Omega. They will be overlapping so we can get some more variation.

 

It is true to say NIR is pretty bleached out, but SWIR has more variation because several important transitions related to hydrogen bonding take place in SWIR, so sugars and water and other carbohydrates are likely to be interesting. (That’s why people do near infrared spectroscopy. The spectroscopic definition of NIR goes out to like 3000nm I think.)

Link to comment

I think I may have struck it lucky with my IR tri-colour filters as a result of having made a stupid mistake ...

 

For my 800-900 range I had bought a Midwest Optical BP850 48mm filter from discount-optics on ebay - only to realise too late that I should have got the narrower-band BN850. So I went back to discount-optics and ordered a BN850 - there was only a 37mm version listed, so I took that. Amazingly, I got a message from the seller saying he was about to list some larger versions and would I prefer one of those (same surprisingly low price of $20) - so I'm getting the far more useful 46mm.

 

But better still, I asked what he (or she) had for 700-800nm and for 1000nm longpass (as an alternative to the 920 longpass I've already ordered). He was able to offer a 43mm MidOpt BP735, which is perfect for 700-800 if combined with an R72, and a 67mm MidOPt LP1000 - again for $20 each. (Actually, I think the LP920 will be better than the LP1000).

 

These filters will give me the following spectral coverage (which has CMOS sensor sensitivity factored in), which will be almost exactly what I wanted (once appropriate exposures have been applied to equilise the curves):

 

post-245-0-02522100-1579210052.jpg

 

I hadn't heard discount-optics mentioned before. Does anyone know them? His filters are mainly MidOpt and LifePixel. The MidOpt ones are extremely cheap - I wouldn't have been able to afford them at normal prices.They're classified as "Pre-owned" or "Opened - Never Used".

Link to comment
These filters will give me the following spectral coverage (which has CMOS sensor sensitivity factored in)

This doesn't make sense with the curves above? The sensor gets less sensitive as you go toward the longer wavelengths, but the filter transmission goes down for the last filter? That last one will pretty much have the 900-1000nm part only, but with much lower peak transmission than the other two. You need it to go the other way.

Link to comment
It's just a tool I build for myself using Excel and VBA. I've entered the data from filter data sheets, and then I can choose multipe filters to display graphs for, and can see the effect of combining 2 filters.
Link to comment

Same technique as above, but this is with a salted caramel cookie. I changed the filters to 780nm, 1064nm, 1500nm long pass (to 1600nm where Triwave's QE dies). The caramel chip seems to turn red with this combination!

post-94-0-50716700-1579235413.jpg

 

post-94-0-53319100-1579235725.jpg

Link to comment

Your sugar crystals also look pinkish.

Really cool.

But don't eat a cookie if it turns all bluish black like that. It may have turned.

Link to comment

I wonder what the thought process was that led you to photograph a cookie. Anyway, it's interesting that the sugar crystal are pinkish - implies there is a drop in reflectance as you cross the border from visible to NIR, but that reflectance picks up again as you go further into IR.

 

That is nifty. Where is the sensor gain data from? Did you measure it?

 

No - I just used a generic graph I found on the Internet. I don't have the technology to do it properly. But if anyone has data for a Sony A6000, I would love to have it.

Link to comment

I wonder what the thought process was that led you to photograph a cookie. Anyway, it's interesting that the sugar crystal are pinkish - implies there is a drop in reflectance as you cross the border from visible to NIR, but that reflectance picks up again as you go further into IR.

1) That crystal is probably salt, not sugar -- it's a salted caramel cookie.

2) I suspect that red may either be reflection from the red caramel or possibly my white balance is a bit off?

3) As for why a cookie, take a look at the IR absorption spectra for sugar (in various forms):

post-94-0-88683800-1579293020.jpg

 

I thought that might lead to some interesting colors for baked goods, and so it seems...

Link to comment

One thing I'm really curious about is what will happen if I do similar tri-band images for fall leaves.

post-94-0-12306700-1579293669.png

https://webbook.nist...7235407&Mask=80

 

I really want a camera that can do out to 3000nm or so, someday. I bet you can do some great color images out there, given how the spectra start going all over the place in that region.

Link to comment

I really want a camera that can do out to 3000nm or so, someday.

Scanning photodiode camera?

 

Also, in that region you have to discern between emitted and reflected light.

Link to comment

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...