Jump to content
UltravioletPhotography

Composite UV imaging using multiple filters


Recommended Posts

Inspired by a couple of threads on here (namely Bernards great examples, looking at using 3 filters to cover the UV range - https://www.ultravio...dpost__p__30286, and Andys post about the limited range of colours in a typical UV image - https://www.ultravio...dpost__p__30024), I've been thinking about using multiple filters to cover the UV range, in combination with a monochrome camera, and assigning the image from each filter a different colour. Then the images would be combined to make a composite UV image with a wider range of colours, than one done with a single wide band UV filter, and converted camera with the Bayer filter still present.

 

This thread will cover my work on this, as it may be of interest to some on here. It's going to be a work in progress, as I have discovered it's a lot more complicated to implement than I initially thought, and I need to do a lot more work on it to get it to where I am happy.

Link to comment

Choice of filters

I've gone down two potential routes for this.

 

Firstly, 3 relatively wideband filter, covering as much of the range from 300 to 400nm as I can. These are surplus ones from Omega Bob. Unfortunately he didn't have three which had as good an overlap as the ones Bernard was using. So I'm not sure how well these are going to work out. They are currently with family in San Francisco, and I'll be getting them towards the end of November. I shall update on these as and when I know more.

 

Secondly, use narrow band, 10nm FWHM filters from 300 to 390nm. I initially though Asahi for this, however total costs and the additional factor of import duty when importing from Japan steered me away from them. I ended up buying some from Edmunds and some from Thorlabs. I got 300nm, 310nm, 320nm and 330nm from Edmunds, and 340nm to 390nm in 10nm intervals from Thorlabs. All of these are supposed to be atleast OD4 to 1200nm, except the Edmunds 300nm one. This one is an outlier, and I'm not sure exactly how I'm going to deal with that just yet.

 

A quick set of transmission spectra for the 310nm to 390nm ones, measured on my Ocean Optics FX spectrometer up to 800nm. Data is available from the suppliers up to and in some cases beyond 1200nm for them.

 

post-148-0-16808600-1570532787.jpg

 

post-148-0-21610200-1570532784.jpg

 

The Edmunds filters (310nm to 330nm) offer more transmission than the Thorlabs ones (340nm to 390nm), and were thinner, but were more expensive.

 

They look to be giving good overlap throughout the UV range from 310nm to 390nm.

 

I ended up mounting these in Thorlabs 1" tubes. The filters are 25mm diameter, and just drop straight in (the tubes have retaining rings to hold filters).

 

post-148-0-69320700-1570536842.jpg

 

Mounting these to a 52mm filter thread for the lens took a little more thought. Thorlabs do an adapter to go from these tubes to 30.5mm. And I bought a 52mm to 30.5mm step down ring from eBay. However the two 30.5mm threads have different pitches. So they fit together, but the threads bind after about half a turn. It's enough to mount them without a light leak, but not ideal.

 

Choice of lighting

 

The original aim of this work was to use a monochrome converted camera in conjunction with the Rayfact 105mm UV lens to image between 300 and 400nm. The Monochrome converted camera would give the best sensitivity at the shorter wavelengths, and a more linear response as a function of wavelength than one with the Bayer filter still in place.

 

For lighting, I wanted something which had more intensity at the shorter wavelength end (to try and offset the camera response, and enable the exposures at the different wavelengths to be as close as possible). Given this, I used my UVP 302nm transilluminator, which has 6x 8w 302nm fluorescent tubes, and emits loads of UV. The spectra of the lamp is shown below;

 

post-148-0-92502400-1570623247.jpg

 

While this does emit loads of light, there is very little after 370nm. As such I decided to use this for between 310nm and 370nm, and not to try doing 380nm and 390nm with it. This may change in the future, if I swap one of the 302nm tubes for a 365nm one for instance.

Link to comment

Calibration of exposure

 

The first thing I wanted to do here was to calibrate the exposure times required at the the different wavelengths, such that the same (or as similar as I could get) exposures would be obtained for each image.

 

For this I used a Labsphere Spectralon diffuse reflectance standard (40%), and stuck it where I'd be doing my images. The lights off, other than the 302nm transilluminator, I took a series of exposures of different shutter speeds with each filter, with a constant ISO and f stop on the lens.

 

Equipment - Nikon d800 monochrome conversion from MaxMax, lens Rayfact UV 105mm f4.5 (at f32), ISO3200.

 

I took the images as RAW files and opened them in RawDigger as RAW composite files, getting the Red, Green, Blue and Green 2 values within the diffuse reflectance standard part of the image. These values were then averaged, to get the camera response at each wavelength and each shutter speed.

 

These were then plotted;

 

post-148-0-05817700-1570634275.jpg

 

Looking horizontally across the graph I then decided on a set of f-stop changes required when moving from one filter to the next, to keep the overall exposure of the diffuse reflectance standard approximately constant. These were;

 

Moving from 310nm to 320nm -1/3 stop

Moving from 320nm to 330nm +1/3 stop

Moving from 330nm to 340nm +1 and 2/3 stop (big jump here driven mainly by the much lower filter transmission of the Thorlabs filters for 340nm and above)

Moving from 340nm to 350nm +2/3 stop

Moving from 350nm to 360nm +1/3 stop

Moving from 360nm to 370nm +1/3 stop

 

In theory, using this approach, the exposure for all the images should be within 1/3 stop of each other. Some of the differences are less than 1/3 stop, but this is meant to get things approximately right.

 

As expected when I started testing, there was essentially no light at 380nm and 390nm so this source isn't suitable for the full range of 310nm to 390nm. I shall need to rethink what to about that. Perhaps replacing one of the 302nm tubes with a 365nm one. Or I may even move away from that light completely and move to flash with quartz tubes.

Link to comment

Initial images at each wavelength

 

Some initial images at different wavelengths with this setup. Far from ideal, and I need to back to drawing board on lighting (I think flash is needed), but early days.

 

Subject was some yellow flowers from the side of the road (nothing flowering the garden at the moment). Basic colour image of them;

post-148-0-90747700-1571407665.jpg

 

I then took 7 images of the flowers and a 40% diffuse reflectance standard using the 302nm light, and a Nikon d800 monochrome camera (ISO1600, Rayfact UV 105mm lens at f32). The images were imported into Darktable, and had a base curve applied (but no attempt at white balance), then saved. I then cropped them in Photoshop, and resized in Xnconvert. Here they are;

 

310nm (4s exposure)

post-148-0-35770500-1571408643.jpg

 

320nm (3s exposure)

post-148-0-90715900-1571408647.jpg

 

330nm (4s exposure)

post-148-0-64595200-1571408650.jpg

 

340nm (13s exposure)

post-148-0-22561000-1571408653.jpg

 

350nm (20s exposure)

post-148-0-93952200-1571408655.jpg

 

360nm (25s exposure)

post-148-0-50358200-1571408658.jpg

 

370nm (30s exposure)

post-148-0-25326200-1571408661.jpg

 

Some initial thoughts;

 

Not huge differences in the flower appearence at the different wavelengths. Biggest difference was in how the glass vase behaved - black at 310nm and transparent at 370nm.

 

Also the flower moved between the images - not surprising being bombarded with short wavelength UV for so long, although I did put a cover over the light between images.

 

The 360nm image is soft - lorry went past during image capture. Need to use flash and get shorter exposure times.

 

The 370nm filter shows some weird 'ring' artifacts, not seen with the other ones.

 

Exposures reasonably well balanced based on the data from the calibration chart.

 

 

Not coloured them or tried to align them in Hugin yet.

Link to comment

Good luck with this - looking forward to seeing your results.

 

There may be one practical issue which comes to mind - when overlaying the images I have to painstakingly resize the 345nm image and then the 380nm image to get a good registration of all three. This is because with the Cassar S I have to re-focus for the 315nm image - but even the 345nm and 380nm images don't quite register: may be something to do with differences in filter thickness? (And with the El Nikkor, I have to re-focus for 345nm as well.) The good thing, though, is that it makes you concentrate on a smaller number of interesting shots rather than on quantity.

 

Anyway, this will be a lot more bothersome if you have 9 images to overlay. May not be so bad if your Rayfact lens really don't show any focus shift.

Link to comment

Jonathan,

This will be great to see when you get some images.

As for lighting, I would recommend using your 302nm lights. So that the light emmission counter balance with your camera sensitivity. This should help to get the best overlap with your selected filters.

Two 15 W G15T8e bulbs should work well to illuminated most subjects. Just stand clear and don't forget to wear yor googles.

 

I forgot earlier today with a flash test I was setting up. At least I have the habit of counting bursts and closing my eyes. So not too bad.

Link to comment

Thanks for the feedback. I'll update it when I have time to get everything into the thread.

 

Bernard. For my initial experiments I cheated - I used f32 on the lens, and didn't worry about refocussing. The Rayfact is pretty good for that anyway, and doesn't show much focus. Overlaying the images will be a bit of nightmare. I found when changing the filters, I moved the camera slightly. Also my test flower wilted very slightly over the time it took to get everything. Also I have wooden floors, and a busy road outside - long exposures and the occasional blurry image. Some of the many, many challenges to getting this to work.

 

David. Yes, my initial though was to use 302nm lights, as I have a 302nm UVP transilluminator with 6x8w 302nm tubes. Very dangerous though.

Link to comment
JMC - again, use Photoshop's Align command for the images. It will do the work for you so long as you get it reasonably close.
Link to comment

Andy, I have never worked with Layers in Photoshop (I'm assuming that's the approach - multiple layers then align), so I'll need to figure out how to do it first. My version of Photoshop is a little clunky now (version CS), and I'm not the most technically capable person when it comes to software....

 

Edit: I'm not sure mine has an auto align function, only to the left or right sides, or top or bottom. Oh and Layers are complex to deal with if you've never tried it before :).......

Link to comment

JMC - again, use Photoshop's Align command for the images. It will do the work for you so long as you get it reasonably close.

 

I use GIMP rather than Photoshop. The functionality is usually equivalent, but the GIMP Align function doesn't seem to do anything like this. However, your post prompted me to look at what other software I have, and after a bit of playing around it looks as though Hugin (another bit of freeware) will do the job very easily. I'll try the whole process out this evening, and let you know how I get on. It will be a real boon (to both myself and JMC) if it works OK - no need to use Phtotoshop or get to grips with Layers (which I have used a bit, but am still not very competent in).

Link to comment

Well, Hugin works a treat!

 

Once I've got the three component files as .tif (having removed profile and white balancing from the RAWs using RawTherapee) I follow this process to get Hugin to align the images:

·
Put the three .tif component files into the alignment directory (Desktop\UV Colour\Align in my case)

·
Open a cmd.exe window

·
Cut & Paste this line into cmd.exe to change to directory holding images to be aligned (include the quotes):

cd "C:\Users\Bernard\Desktop\UV Colour\Align"

·
Run the Hugin alignment command (include the quotes):

"D:\Program Files\Hugin\bin\align_image_stack" -a Aligned .\*.tif -m

·
The output will be 3 files with names of the format
Aligned0000.tif
in the order of the input files (which will be 315nm, 345nm, 380nm if the input files are named in my standard way).

·
Move all of these files out of the alignment directory in preparation for the next alignment run.

 

The -m switch at the end of the command is essential to automatically rescale the images to account for re-focussing, different filter thicknesses, or whatever.

The .tif files that are output can then be used to create the composite TIFF (in my case by dropping the three images into GIMP, changing mode to to greyscale, and then using Color>Components>Compose), followed by the final white balancing in RawTherapee.

Not only is this faster than manual rescaling and registration, it is more accurate (as my patience runs out when doing this manually): no matter how much I enlarge the final image, I can't see any colour fringing.

In the above notes I am using 3 colour separation images, but the process will work with larger numbers.

 

PS: JMC, although I have a concrete floor downstairs in the kitchen, for this trial I was upstairs in a room with carpets and floorboards, and a main road just outside the house, like you. That didn't cause any problems in terms of vibrations: admittedly I'm using flash, but for the 315nm shots I need to use 6 flashes which takes about 30 seconds, and there is no noticeable camera shake.
Link to comment

Ooh, that seems handy. Since that's command line, I could call that from MATLAB even...

 

JMC - go with Hugin method. I would bet it's better than PS anyhow.

Link to comment

... Also my test flower wilted very slightly over the time it took to get everything. ...

 

A problem I am familiar with from doing macro stereo shots using focus stacking. Could be up to 150 exposures for one stereo pair. In the time to take all these I experienced some flower wilting - but I also came to realise how much flowers move (rather than just droop)! The stamens/stigma/anthers (never can remember which is which) often seem to wave around over a period of minutes. On one occasion I was working with a magnolia flower, and I could see its petals twitch every 30 seconds or so.

 

As you will be making 9 exposures, this will be a problem you encounter with certain types of flower.

Link to comment

Start Botany 101

  • A stamen ends in an anther. This is the male part.
  • A style ends in a stigma. This is the female part.

 

End Botany 101

Link to comment

Start Botany 101

  • A stamen ends in an anther. This is the male part.
  • A style ends in a stigma. This is the female part.

 

End Botany 101

 

Birna if you start a course I may attend. The only only Biology course I droped was plant physiology.

Link to comment

Thank Bernard, I shall take a look at the Hugin software then. After some time playing with Layers in PS last night, I still haven't got my mind round how they are working. Any alternative will be good.

 

My exposures where anything up to 30s at ISO1600 using that 302nm light. I think flash will be the way forward (vs the continuous fluorescent light), although that in itself will provide challenges about getting the same exposure at each wavelength.

Link to comment

I think flash will be the way forward (vs the continuous fluorescent light), although that in itself will provide challenges about getting the same exposure at each wavelength.

 

I need 50x as much exposure for 315nm as for 380nm. So I ramp up the ISO (typically from 200 to 1600) and use 6 flashes (4 flashguns for each): but that still takes about 30 seconds!

Link to comment

I've got the Hugin re-scaling/alignment process running very smoothly now, using a single command line which I cut & paste into the cmd.exe window,

 

The generic structure of the command line is:

 

"{Pathname to Hugin Alignment executable}\align_image_stack" -a "{Pathname to output folder}\{Prefix for Output filenames}" "{Pathname to folder holding the input files}\*.tif" -m

 

The quotes are required.

 

You can use .jpg instead of .tif (but the output will always be .tif).

 

In my case, the command line looks like this. I am using the same folder for input and output. The output filenames are of the format Aligned0000.tif :

 

"D:\Program Files\Hugin\bin\align_image_stack" -a "C:\Users\Bernard\Desktop\UV Colour\Align\Aligned" "C:\Users\Bernard\Desktop\UV Colour\Align\*.tif" -m

Link to comment
Thanks for that. It seems very convenient. I have Hugin and didn’t know it had a command line interface. Bernard, never use JPG for this purpose- it alters colors. Most of the information that gets thrown away is in the chromaticity, I’ve found (you can see it when you look at L*a*b channels for a JPG. The a and b are very blocky.)
Link to comment

Updated with some initial images at the different wavelengths. Not tried colouring them yet, or aligning them in Hugin. Bit busy with work, and not had chance to do more with it. Will do more when I get chance.

 

Not hugely happy with them yet - more to do to make them better.

Link to comment

Couldn't resist having a go with your images.

 

Not sure how you're planning to colour them, but I combined 310, 320, for the blue channel; 330, 340, 350 for green; 360, 370 for red.

 

post-245-0-36760000-1571433812.jpg

 

The white balance is pretty good without any additional work.

 

The red glass is sort of what you'd expect - stronger absorption at shorter wavelengths with longer wavelengths being transmitted.

 

Looks like your flowers were moving around a bit, though! And there's some colour fringing at bottom-right and around the edges which I don't understand.

Link to comment
Bernard, that is brilliant. I had no idea how to do that. I was planning on giving each one a different colour, but that's a great start. Yes the flowers did move a bit, and I think there was a bit of tripod movement too. The floor is quite springy. Overall, plenty of room for improvement.
Link to comment

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...