A Calibrated Gray Scale for Forensic Ultraviolet Photography

[nfoto]

Interesting approach which holds promise indeed.

 

Your illustration picture has a slight cold colour cast, though. Perhaps not a big surprise considering the spectrophotometer graph.

[JMC]

Thanks Bjørn. I'm not sure I would read too much into the colour cast of the photo though, it was taken with the light coming through a window in bright sunlight, with the camera on full auto (including the white balance), although yes it would tie in with the reflectance graph data.

[Andrea B.]

Cool to see one of the recipies actually cooked up! Thanks for showing us.

********************

 

A question: how exactly are you going to use your reflectivity standards between 2% and 20%?

I'm asking this question for a variety of reasons one of which is for my own education. "-)

 

Two possibilities come to mind:

1. Reflection standards used to compare UV reflectivity of photographic subject to known UV reflectivity.
   
2. Reflection standards used as proxies for "brightness" and used to adjust photo in the converter.
   

Given that you have mentioned elsewhere making UV photographs involving sunscreens, I'm guessing your use for a reflectivity standard would be #1?

 

From experience I know that #2 is very difficult because of the way raw file conversions are performed. Midtones are lifted and the 25-50-75% (as just one example) reflectivity from standards does not translate well to a 25-50-75% brightness in an HSB model.

 

If you use of reflection standards is #1, then wouldn't you need something very smooth and diffuse so that you do not reflect more light at one angle than another and perhaps misjudge the amount of reflectivity from your UV subject. (Which also reminds us that we need to illuminate the subject very evenly with the UV.)

 

Do you think your spectral measurement might have been flatter (and smoother) if you sanded and smoothed the surface of your target?

[Andy Perrin]

Andrea, usually diffuse reflectance is improved by increasing roughness, not going the other way.

[JMC]

Hi Andrea. Yes it's number 1 of your two options. When I've looked at some of my samples on the UV Vis spectrometer they are typically reflecting between 5% and 15% of the UVA. I want a way of estimating the degree of reflection directly from the photos, hence creating some standards.

 

I may try sanding them, but the main issue is dust. Sanding creates a dusty surface and if I get any of that in the UV Vis device it will be a very expensive mistake. When I measure the samples on the UV Vis they are upright so it'd be very easy for dust to fall into the integrating sphere. Perhaps sanding and then wet wiping will work, we shall see.

 

I made another one today to try vibrating it to get the air bubbles out, which worked very well. But I think the method still needs a bit of work to optimise, as although though the surface was smoother it looked a little heterogeneous in terms of reflectance.

[Andrea B.]

Slightly overfill the jar lid and use the flat side of a knife to scrape across the top to give a smooth surface. That's how we ladies level flour when making cakes. (Please don't ask me when was the last time I made a cake!)

 

Why are the air bubbles occuring when you mix in the water?

[JMC]

I'll give that a go thanks. I think the air bubbles were from one of the powders not wetting out fully, so I was getting little pockets of dry powder in the first attempt. Between better mixing and a bit of vibration to settle everything, today's attempts are getting better.

[JMC]

Slight update on the quest for a 'cheap' grey scale calibration standard. I tried making a few mixtures based on the literature, and pouring them into small (approximately 4cm diameter) jar lids to set. To be honest it did not go well - 3 of the 4 cracked badly on drying.

The one that did not crack on drying, I measured on the UV Vis for reflectance and it was more variable than the one I 'eyeballed' the composition for a few days ago. I think the new lids are too big, with too large a volume for the material, so that when it dries there is a lot of shrinkage which is causing the cracks. I'm going to try partially and fill them with something that wont dry first, to reduce the volume, and then redo it. I'm also going to try making a small card mould, sort of like a miniature picture frame, and try pouring it into that to dry as well. Andrea, you're suggestion of overfilling and then scraping off works very well and I will stick with that for some further method development work to try and optimise.

 

Second experiment - mixing carbon black into matt white paint. This was a real suck it and see experiment, where mixed 1, 2 or 3 scoops of carbon black into white matt paint to get light, medium and dark grey. No measurements for this (for the amount added). I painted this on to card, and masked round the edges, as to not let any of it get close to the window of the UV Vis during the reflectance method.

I managed to get reflectance data from this, and while they were obviously different, the reflectance in the UV was very non linear with regards to wavelength - much worse than the MgO/plaster/carbon black mixture. I'm guessing this is strongly driven by the chemical composition of the paint. Also the carbon black did not disperse well, and formed lumps it was very difficult to get rid of in the paint. Overall very disappointing, and the 'mixing into paint' is not something I am going to try refining as a method.

[OlDoinyo]

If you paint something onto the back of a transparent (i.e. quartz or very thin glass) pane, would that give you a more uniform reflecting surface than one just exposed to air? I am assuming that the swatch would be illuminated/viewed through the pane.

[Andrea B.]

Andrea, usually diffuse reflectance is improved by increasing roughness, not going the other way.

 

Yes, you are correct! I mis-spoke.

I was trying to indicate "smooth" relative to the lumps which are showing up in the plaster of paris mixture. :lol:

 

Interestingly, my Spectralon standards, when sanded underwater to clean them up, feel so "smooth" and are so "smooth" that water beads up on them and rolls off. Yet they reflect incident light diffusely. So the "roughness" is exceedingly "fine grained". Language is failing me here. Hence all the quotes. Halp!!!! :rolleyes:

 

*****

 

Jonathan - kudos for your persistence. If you do get this worked out, then you might just have some eager customers wanting to buy a pot or two of your pigments for their own use.

[JMC]

Hi OlDoinyo, I am wary about making a standard which has something over the front of it, even if that something is UV transparent. To me it always raises that question of "what if anything is the cover changing about the light transmission"? I have a couple of things to try first before I go down that route.

 

Hi Andrea, Interesting what you say about sanding your standards to clean them. Does that have to be done in a very controlled way (size of grit for the abrasive paper for example)? I would have thought the degree of roughness would have impacted the amount of reflectance. Spectralon is sintered PTFE, so has a very low surface energy. As such it will repel water anyway. If it's roughened (even on the nano scale) the degree of repellence increases, and very fine grained roughening will have a marked effect on making it more water repellent. I will keep trying to improve on the method for making these as time allows - it's a bit of a weekend job for me at the moment, so posts may be more 'occasional' moving forward.

[Andrea B.]

User Guide: Spectralon Reflectance Standards Care and Handling Guidelines

From this PDF:

...the material can be cleaned by sanding under running water with a 220-240 grit waterproof emery cloth until the surface is totally hydrophobic (water beads and runs off immediately). Blow dry with clean air or nitrogen or allow the material to air dry.

 

 

The thing about Spectralon standards is that they are really not for field use in UV photography. Which is exactly where I've been using them. So they have been dropped or blown over, become dusty or pollinated and have gotten little scratches. Every now and then I do a bit of sanding as suggested to tidy them up. I'm quite sure my set have all drifted off their original calibration. (See footnote!)

 

But I don't really need that accuracy. (Or is it precision? Sorry I always mess up accuracy/precision.) For example, if the 50% reflective standard is now reflecting at 51%, that's OK because for the way I use the standards, having "midtone" is good enough. And of course for simple white balancing, it's ok if the 99% standard is really at 98% or whatever. The black standard has deteriorated the most - probably because it is less Spectralon and more pigment.

 

Do I really need those Labsphere standards just to record a few botanical UV-signatures? No, probably not. :D But I have really enjoyed learning from them. I'm the happiest lady in the world because I get to play with this UV stuff and make little experiments and talk to other folks who enjoy the UV world. It's endlessly fascinating. B)

 

*****

 

P.S. I should add that the reason my standards may have "drifted" would be due to picking up certain contaminants -- like skin oils or similar -- which could not be removed by rinsing/sanding. Spectralon itself isn't going to change its reflectivity with age.

[JMC]

Bit of an update on these. I've been playing around with the pigment method of making the standards, and seem to have found a way now of making them dry without cracking, and producing a nice relatively even finish across the surface. I made up some mixtures of the MgO, Carbon black and plaster of Paris in different ratios, cast them and let them dry. Run as before on the UV-Vis I have access to for % reflection between 250nm and 800nm. I prepared 5 new samples (recipes 1, 2, 3, 4 and 6) and had 2 older ones from about a month ago (recipes 5 and 8). The recipes were meant to go from most reflective, with recipe 1, through to least reflective, recipe 8. The whole spectrum graph is shown below;

 

At first glance this doesn't look ideal, given the non linearity especially for 1 and 2. However for my needs, ideally I want standards in the 0-10% reflectance range, and in the UV (320-390nm). So replotting with that in mind gave me this;

 

This looks more promising. The standards look different to each other and relatively flat in the wavelength range of interest. 4 of them are very close together though, so it remains to see whether photos of each under standard lighting will look different during image analysis. Given I have previously run a sample of Semple 2.0 black paint (about 2.5% reflectance in this region) and the Gretag Macbeth colour chart black (which is neutral in this region and about 3.3% reflectance) I now have a few standards to play with for imaging.

[JMC]

Got a photo of the calibration standards today. Monochrome EOS 5DSR, Asahi UAT 85mm lens, f5.6, ISO800, 1/25th second, bright sunshine about 11:30AM, Baader U filter, saved as monochrome jpeg in the camera and not further modified. I included the MgO/carbon/plaster sample I had, and the Semple Black 2.0 paint I've looked at before as this has a low reflectance. Image given below;

 

Samples 1 and 2 are quite over exposed, but as mentioned before it's really 3-8 and the black I'm interested in, as I'm primarily looking at 0-10% reflectance in the UV with my samples. So I took the image and cropped out the centres of each of the samples in Photoshop, and ran each of these through Image Fiji (Image J, https://imagej.net/Fiji) to calculate the histogram greyscale score. This gives a value from 0-255 (0 being black and 255 being white) along with a standard deviation in the grey value. I then plotted the greyscale score from the image against the percentage reflection from the UV Vis analysis;

 

Overall, the samples tracked as expected - lower reflectance as measured on the UV Vis corresponded to darker image from the camera (lower greyscale score). It looks like under these settings, the blackest sample (Semple 2.0 paint) is still relatively light in the camera, which was because I over exposed everything. I did this as I'm more interested in the lower percentage reflectance region. What I am surprised at is the ability of the camera to differentiate very similar reflectance samples - it looks like it can differentiate down to about 1% difference in total reflectance. I should stress this was a 'suck it and see' experiment, and I didn't really optimise settings for capture, and it was using sunlight. I expected a non linear relationship between the greyscale score and measured reflectance, but I guess because I am in the middle of the possible greyscale values, that portion of the relationship is approximately linear hence the fit. Bit cheaper than trying to get custom Spectralon samples, but I suppose you do need access to a UV Vis to get the reflectance :)

[Andrea B.]

That is REALLY looking good, Jonathan.Your casts are looking nice and smooth.

 

Remember that your camera is not going to be able to capture the entire reflectivity range of all samples in one photo (under the assumption that brightness can be reasonably equated to reflectivity). Something will always be too dark or too bright. So when using standards like this, it is important to work out a conversion method. (That is a little vague. I should post an example of what I mean.)

[ulf]

I find this thread very interesting and hope to be able to make some UV-gray reference myself for white balancing.

With some of my filter combinations, especially the UV-BG-variants, I have had problems using a PTFE sheet as the R, G and B-channels see very different amounts of light.

To avoid saturation with the PTFE sheet I have had to use very different exposure settings compared to for a normal scene.

 

I did some experiments with the material mixtures posted in this thread, but had had problems with getting an evenly grey surface.

Something in my methods is not good and I also get tiny black air bubbles, even if I vibrate the mixture before and after I have applied it.

 

However I might add an idea of a substrate I think is useful.

I used the recess at the rear side of very cheap UV-filters from eBay / Aliexpress for the reference mixture.

The recess is only around a millimeter deep and the mechanical shape with the threaded metal ring is stable and clean.

 

After ruffing the glass surface on the rear side, I placed the mixture in the recess and scraped off the top to make it even, as recommended by Andrea.

When the mixture is settled and dried it could be protected by a metal filter cap.

I had no problems with cracking.