23 replies to this topic

[Nico]

This article is meant as a brief documentation of how I generate the "false colour representation of the spectrum visible to bees" that has been described in the two previous articles.
I use Lighroom and Photoshop in this example, but there are certainly many alternatives, e.g. GIMP that can be used in a very similar way.
If you have not read the previous two articles in this section you might want to start there:

• Simulation of Bee-Colours I
  
• Simulation of Bee-Colours II

Start with a visible light image and an UV-image converted to black and white that are well aligned:


Open both images in Photoshop:


Activate the visible light image. Go to the right and open the channels panel, there select “split channels” and you end up with a different image for each channel:


Repeat the step above with the steps above with the UV image:


Now you have six independent images. However, those from the black and white UV-image are all identical.

Go to the channels panel and select “merge channels”:


In the window that opens, select “RGB”:


Then select the channels as show. Put blue of the visible light image in the green channel, and green in the red channel. A channel of the UV image goes in the blue channel (it doesn’t matter which, since we have used a black and white conversion. I usually use blue, as shown here).


Then select “ok” and you get an image similar to the following:


Alternatively, you can leave blue and green of the visible light image in their respective channels and put UV in the red channel, as shown here:


This will result in a different false colour representation of the spectrum visible to bees.


Since I have published the first two articles of this series and now I have been experimenting with some filters that transmit roughly the spectrum visible to bees. This approach allows generating similar false-colour representation with a single exposure.
Some examples can already be found here: http://www.ultraviol...photographycom/. More on that topic in a future publication.

Published, 5 April 2014

[DaveO]

Thanks Nico,

Now I see how it's done. It made my head hurt trying to figure it out. Now, all I need to do is figure out a simple way to align the visible and UV images, it's easy if you treat them as layers but my head is starting to hurt again.

Cheers,

Dave

Edited by DaveO, 06 April 2014 - 03:58.

[colinbm]

Fantastic work Nico
I do prefer using filters, but I haven't got the green to come through the UV yet.
Col

[Nico]

@Dave
I try to capture two shots that are "perfectly" aligned, rather than trying to align them in PS. If the camera is steady on a tripod and the flower doesn't move it works, sometimes. Unless you are in the studio, you need some luck, obviously. However, I managed many times in the field, so far.
You can achieve the very same effect with layers if you first merge each black and white layer with a layer of the respective RGB colour and then merge the three resulting layers together with the appropriate algorithm. I’ve used that way too, but it seems more tedious to me. However, you might have more align options … I’m sure that PS gurus will have many more elegant ways to accomplish this.

@Col. - Thanks!
You need a filter that transmits UV, blue and green and you need to supress the longer wavelengths. It is helpful if there is only a small amount of blue and green transmitted, since the camera is less sensitive to UV.
From the options that I have tried a stack of Schott UG-5 and S8612 (or BG39), 2mm each work best, especially if UV is further boosted with a flash as shown in the image below. I have done a bunch of experiments and hope to publish them in a more formal way within the next 2-3 weeks.

Cheers,
Nico

This image of Ranunculus ficaria was captured with UG-5 plus S8612 (2mm each) + a UV-flash (only slightly tweaked afterwards):

Image reference: CHA_P1120949_bee

Edited by Nico, 06 April 2014 - 09:24.

[colinbm]

Thanks Nico
I have some more filters now incl a S8612, so I will try again with the UV-Green filter + S8612, as I may have been getting a very slight IR leak before.
Thanks for the reminder
Cheers
Col

[Alex H]

Is your UG-5 and S8612 stack similar to what Klaus Schmitt uses for his work, Niko?

[Nico]

Hmm,
I don't know for sure, since Klaus didn't tell me, and I couldn't find a documentation on his blog, either.
However, since there is a limited number of filter glass types on the marked and since the resulting images look quite similar, I guess the answer is most likely "yes" ...
There's also an ebay seller called "uviroptics" who is offering several filters and also stacked combinations (One called Bug U5).
He has been very open about the composition of his stacks when I got in touch with him. - I bought some single filters from there.

Best, Nico

[Alex H]

OK, thanks. I have some questions about these UG-5 stacks and the resulting pictures, but I will wait until you publish your descriptions before asking them.

[msubees]

thanks for posting the details. my old photoshop cs3 (=9.0) is perhaps too old. it does not have merge channels option. and the channels are automatically split when I click on channels. I tried to use channel mixer to see if i can reproduce the same thing but i could not. I do not want to really to pay monthly for the CC version but can get a CS5 on the web for about $75.

[Andrea B.]

You can make channels without a channel mixer, Zach. This old-fashioned way is a bit tedious, but it is correct. I only have Photoshop Elements, so I have to make my channel stacks 'by hand'.

RGB Channel Creation the Hard Way.

• Open the photo file in PS.
  
• Add a new layer over it.
  
• Fill the overlayer with Red (255,0,0).
  
• Change the blending mode of the overlayer to Multiply.
  
• Merge the two layers⁽¹⁾.
  
• You now have the Red channel of your photo.

Repeat and rinse to create the Green channel and the Blue channel for your open photo.

Do this for both your UV and your Visible photo so you can mix layers from each to make a composite photo.
(You will have 6 channel layers.)

(1) In the layer merge step, I typically do a PS "stamp" operation to create the new channel
so that I can continue to use the open photo for creating the next channel in the same stack.
On the mac this is accomplished by simultaneously hitting the Command/Option/Shift/E keys.

Composite Channel Stack

• Choose one R, one G and one B from your 6 layers. (2 UV + 1 Vis or 1 UV+ 2 Vis, whatever.)
  
• Change the blending mode to Difference for each layer.
  
• If needed, align the layers so the Visible and the UV match up.
  
• Merge the 3 layers to get a new "multi-spectral", composite photo.

Math combinatorics question:
How many composite difference stacks can you make choosing 1 each from the 6 R, G and B layers where you may not repeat a channel colour?
(This was on the SAT in 2004.)
.
.
.
j/k

[msubees]

Andrea, thank you! will try next time.

[Nico]

@Andrea
I'm not so sure why I would use the three different channels of a UV capture, rather than converting it to black&white which equalizes all three channels.
The combined information (bandwith) of the entire UV capture is more equivalent to a singe RGB channel of a visible light RGB channel, I think.
Also, since we can take red out of the equation (assuming that we are simulationg thespectrum visible to bees and other pollinators) we have finally three channels which makes the number of combinations much smaler (3x2x1 = 6). In prinziple every combination provides a (valid) false colour representation. However, I find that it makes more sense to use either "UV, blue, green" or "blue, green, UV" than any of the other two combinations. The first one keeps the wavelengths in order, the second one keeps the false colours as close as posible to the human colours. I think it is more a matter of personal preference than a scientific decision ...

Best, Nico

[Andrea B.]

Primarily I wanted to point out to Zach how one could proceed without having a Channel Tool at hand.
Also, we have experimented over the years with multi-spectral stacks of various compositions. So my generic channel stack instructions permit all possible combinations so that one can mix & match various UV, IR and/or Visible frames.

For your Bee Vision work, of course you have chosen the combination to best reflect your model. It does seem appropriate and useful. And we all enjoy your results !!
It is great to have your 3 part essay/tutorial here on UVP. Once again, thank you.

**********

BTW, one of our Nikongear members, Ben Lincoln, wrote some software to automate channel stacking. It is a very impressive effort. You can read about it here and see lots of examples: http://www.beneathth...ace_Breaks.html
It's possible the software might be useful to you for your Bee Vision work?

I moved to Macs and have not subsequently been able to get the software working again.

**********

When you say you use {UV, blue, green} in your composite, aren't you just assigning the UV to the red channel? Putting the greyscale UV image into the red channel produces exactly the same end result⁽¹⁾ as making a red/multiply layer from the UV image in my method above.

Added Later Note (1): There is an assumption there that the original image has been adjusted so that blown channels ("oversaturation") has been dealt with in the original image or later in the new stack.

[Nico]

Thanks, Andrea!
UV in the red channel is one of the two variants that I've been playing with. The other is UV in the blue channel and then blue in the green channel and green in the red channel, as described.

As mentioned earlier I'm currently experimenting with the UG-5 filter stacks. I like the advantage that just one exposure is needed. On the downside, there is some UV component in all the channels. But maybe that's only a minor thing. I still haven't completely made up my mind.
I'll be elaborating a bit on that, soon.
The software that you mention seems very interesting. I'll try it as soon as I find some time.

[Andrea B.]

Nico wrote: As mentioned earlier I'm currently experimenting with the UG-5 filter stacks. I like the advantage that just one exposure is needed.

I don't understand the 'one exposure' reference ?

[Nico]

Andrea B., on 11 April 2014 - 14:14, said:

Nico wrote: As mentioned earlier I'm currently experimenting with the UG-5 filter stacks. I like the advantage that just one exposure is needed.

I don't understand the 'one exposure' reference ?

Well, the images with the "bee colours" that I have shown previously were always based on two images taken with different filtration (UV & VIS).
Since the UG-5/S8612 stack transmits UV plus some blue and green, it transmits roughly the spectrum visible to bees. So, no combination of exposures with different filtration is needed to generate images like the one that I have posted above. That's what I was trying to say. - Does it make more sense, now?

[colinbm]

I agree too Nico
Col

[Andrea B.]

Thanks, Nico.
Looking forward to a U5+S8612 thread in Tests & Techniques. Of all the UV filters, that is the one which I don't have. "-)

[nfoto]

Then, you should address the issue of channel/spectral band balancing, which is not trivial.

[ahrneely]

I don't know if either of these essays are of interest to your simulation of bee vision. We have a guest who stays at our resort each May who had just completed her doctoral thesis on bees and I asked if she had any information on bee vision. She emailed me these two PDFs. They are a bit over my head, but hopefully someone finds them interesting.

Attached Files

•  Bumblebee_vision.pdf   1.08MB   559 downloads
•  Honey_bee_vision.pdf   966.15K   436 downloads