Jump to content
UltravioletPhotography

Thoughts on color correction in UVIVF by subtraction rather than white balancing.


Recommended Posts

I first mentioned this possible alternative to adjusting white balance for UVIVF at the end of the "frozen outdoor" thread, https://www.ultraviol...frozen-outdoor/

with an example inspired by color processing in astrophotography that compensates for light pollution by subtraction. I hope it can inject some fresh thoughts into the discussion - I am not trying to provide any final answers here. While putting together the examples in this write up I realized that it was more complicated than I initially thought.

 

First let us consider the hypothetical situation where we have a light source that is perfectly filtered to only provide UV and no visible light. Also assume that the camera is perfectly filtered to only be sensitive to light in the visible part of the spectrum. The room where we perform our captures is perfectly dark and the only object that shows any fluorescence is the one we want to photograph. How should the camera be color corrected? I would suggest that since we are working with visible emitted light the camera/lens/filter system should be color corrected for a daylight light source. This is ideally done by color profiling with a standard target and a daylight light source. If we have an unmodified camera and we are able to filter out UV without applying a too yellowish filter an in-camera daylight white balance setting might be a reasonable approximation to what colors a daylight adapted eye will see. (Camera manufacturers are pretty good at factory profiling for daylight light sources these days). In astrophotography, daylight WB is typically used as starting point with a non-modified camera when one care about obtaining correct colors.

 

In our ideal UVIVF capture we are now recording only the emitted light from our object of interest. The parallel to this hyppothetical scenario is similar to recording emitted light from stars and other deep space objects in astrophotography with a camera located in space well out of our atmosphere.

 

However in a practical situation we might have several issues:

1) The UV source was not as perfectly filtered as we wanted and emitted a small amount of visible light that is reflected off the object (typically in the blue-violet part of the spectrum).

2) Our filtration on sensor/lens did not filter out UV perfectly and there is some reflected UV contamination.

3) Nearby objects also emitted visual fluorescence that in turn hit our object of interest similar to a visual light source and was reflected back.

4) Our room or outdoor scene was not as perfectly dark as desired but has some background light that was reflected from our object of interest.

 

All of this is contamination of our signal by addition of reflected light. They could for instance cause the image to take on a cold character, but #4 could cause other effects depending on the nearby object. The parallel in astrophotography is the realistic situation that we are located within the earth's atmosphere with light pollution from a nearby city contaminating our signal.

 

The previous discussion here on color correcting UVIVF has seemed to be about white balancing against (very expensive!) UVIVF standards:

https://www.ultraviol...e-fluorescence/

However white balancing applies multiplication factors to all signals recorded, both our signal of interest and the added contamination of the signal under 1-4. In astrophotography it is known that this will cause star colors to become incorrect. It causes color shifts with different signal intensities in the objects of interest. If light contamination has been added to the signal, it should be subtracted to get back to correct colors. https://clarkvision.c...ge.processing2/

Perhaps a similar approach can be taken here with emitted fluorescence. The reflected contamination that was added to the signal under 1-4 can be subtracted instead of trying to adjust with the WB coefficients. In astrophotography this is in practice done by a level adjustment, shifting the bottom of the histogram until the left edge of the histogram peak for the different color channels align so that the darkest tones become close to zero. (Refer to fig 1d in the above link for explanation of how different histogram related levels adjustments work.)

 

This gets a bit more tricky in UVIVF. What could be used is to include a standard that will record reflected light and is not emitting any light. The standard would need to be recorded on site near the object of interest to account for all contamination under 1-4. The light on this standard can then be subtracted from the total signal with a levels subtraction on individual channels as above. Just out of curiosity I brought out my Colorchecker Passport and set it up with some other targets to see what it would look like in UVIVF. I am first showing the visible version lighted by my Niterider Lumina 900 boost LED light at the lowest (200 lumen) setting, My non-modified D7100 was set to daylight WB on all of the following captures:

#1

2018-05-08-0147E-7236-md.jpg.5239c090a5d2ed75b99b92752baf128a.jpg

 

Non-modified Nikon D7100, 55mm f/3.5 micro @ f/8, 0.6s, ISO 100, Nikon L39 filter on lens, Niterider Lumina 900 visual LED source at low setting.

 

Here is the unedited UVIVF version. It was recorded with the D7100 and 55mm f/3.5 micro @ f/8, 30s ISO 100 and UV- painted with a Tank007 TK-566 UV-LED light with a ZWB1 2mm thick filter on the front and an L39 filter on the lens. Interestingly the brightest small white patch near the bottom of the frame is really dark:

#2

2018-05-08-0146E-7235-initial-md.jpg.98443a03d194a19265e1d6fbbacebed3.jpg

 

Non-modified Nikon D7100, 55mm f/3.5 micro @ f/8, 8 sec. ISO 100, Nikon L39 filter on lens, Tank007 TK-566 365nm UV-LED light with a ZWB1 2mm thick filter on the front .

 

 

For this exercise let us just assume that the small bright white patch at the bottom it is not emitting anything and any color on that patch is reflected visible light contamination. A crop is selected on the patch so that the levels adjustment will only show the histogram from this area:

#3

Clipboard01.jpg.4a630fc47254742f21807c52349f7253.jpg

 

 

Before and after subtraction of signals in the blue and green channels on the patch:

#4 - #5

Clipboard02.jpg.cbbe194a64b6b7517e0ff018ef3bee3c.jpg Clipboard03.jpg.e97d1defe329e62682867844cc7c1736.jpg

 

 

This is what the uncropped scene looks like after subtracting what is assumed to be light pollution:

#6

2018-05-08-0146E-7235-subtracted-md.jpg.6a214bf585f71ae9fe0ae742b2bd8e5b.jpg

 

Same capture as above edited.

 

To compare, here is instead a correction by increasing color temperature in the white balance settings. I had to pull the slider all the way up to 10000K. Note that while subtraction above had a pronounced effect on the darker tones, the increase in color temperature here has more effect at warming up the brighter part of the image:

#7

2018-05-08-0146E-7235-WB-colortemp-md.jpg.0fc4f83819047ea53b716ec7ef3a3020.jpg

 

Same capture as above edited.

 

As a side issue it can be noticed that the reflection of the UV LED light in the shiny spoon to the lower right is visible as a violet spot in the above UVIVF capture. I confirmed that in a corresponding UVIIF capture (with my D40x IR-720nm) the spot on the shiny spoon is hardly visible, just included here for the record, Also note that our white patch is now quite visible in UVIIF:

#8

2018-05-08-0213X-4821-md.jpg.dbcc67a1f47edf2d784fff08493fb26d.jpg

IR-modified Nikon D40x (Lifepixel standard ca. 720nm), 55mm f/3.5 micro @ f/8, 30 sec. ISO 100, Nikon L39 filter on lens, Tank007 TK-566 365nm UV-LED light with a ZWB1 2mm thick filter on the front .

 

Below is a detail from another UVIVF capture of the shiny spoon with the UV LED light very close. There is a blue ring caused by lint sticking at the edge and then the violet center. It is reassuring that when viewed through my non-tinted 3M UV- safety glasses only the blue lint is visible, not the the violet center which is only is visible to the eye if I take a quick half-second peek without the safety glasses. As it is unlikely that my eyes or the camera is sensitive deeper into the UV range, this is probably light in the transition between UV and visible light. I feel uncertain how much effect it would have on actual captures. It is not too intense compared to the lint at the edge (The UV light's front glass is also a dark filter that supplements the ZWB1 2mm thick filter that I added on the front):

#9

2018-05-06-2224E-7226-md.jpg.55c24782042ecd45aace459c2cf75856.jpg

 

Non-modified Nikon D7100, 105mm f/4 micro @ f/8, 2 sec. ISO 100, Nikon L39 filter on lens, Tank007 TK-566 365nm UV-LED light with a ZWB1 2mm thick filter on the front kept very close to shiny spoon, moderate crop.

 

The white patch on the Colorchecker that we used for the adjustments appeared slightly blue rather than violet. It could also have been due to possible blue light emitted from other objects out of the frame and reflected off the white target. Whatever source, this is what we tried to compensate for by subtraction on the normally white patch.

 

 

In the second exercise we move back to my scene with the tree stub from the "frozen outdoor" thread. Here is the visible light version again lighted by the NiteRider LED light :

#10

2018-05-08-0247E-7248-Vis-md.jpg.893f786e931e0bcfdbd6b211a0c18c37.jpg

Non-modified Nikon D7100, 105mm f/4 micro @ f/8, 0.5s, ISO 100, Nikon L39 filter on lens, Niterider Lumina 900 visual LED source at low setting.

 

 

As it is not completely dark outside here at this time of the year, let us look at a frame without artificial lighting. Notice that our reference white patch is now much brighter than any part of the tree stub:

#11

2018-05-08-0248E-7249-initial-md.jpg.b9585a20d4ed69b1b1ffe6bcd48e00f3.jpg

Non-modified Nikon D7100, 105mm f/4 micro @ f/8, 30s, ISO 100, Nikon L39 filter on lens, only background light.

 

 

If we want to subtract light so that the tree stub is dark without contamination of light from the night sky but not clipped, we must leave some of the reflected light on the patch. In other words we have a problem compared to the astrophotography situation as reflectance of our reference patch might not be the same as that of the object of interest:

#12

2018-05-08-0248E-7249-subtracted-md.jpg.2b3c1644a47fb7215ec2fb7156e2bff9.jpg

2018-05-08-0248E-7249-subtracted-md.jpg.2b3c1644a47fb7215ec2fb7156e2bff9.jpg

Same capture as above edited. (The remaining dark orange on the tree stub is back light from the window of my cabin - I forgot to turn the lights off.)

 

 

Here is the UVIVF version before any correction:

#13

2018-05-08-0250E-7251-initial-md.jpg.09f5e5b2a752bbb38d25c29f7b06b66a.jpg

 

on-modified Nikon D7100, 105mm f/4 micro @ f/8, 30s, ISO 100, Nikon L39 filter on lens, Tank007 TK-566 365nm UV-LED light with a ZWB1 2mm thick filter on the front.

 

2018-05-08-0250E-7251-subtracted-by-background-capture-md.jpg.eb2951388348f33c9bbdbc881821a8f1.jpg

 

To do the subtraction on the UVIVF capture we have to go "by feeling", due to the higher reflectance of the patch, we cannot make the patch dark which would result in severe clipping of light from the tree stub:

#14

2018-05-08-0250E-7251-initial-md.jpg.09f5e5b2a752bbb38d25c29f7b06b66a.jpg

Same capture as above edited.

 

Applying the adjustment I made on the dark frame without artificial light on the treestub above to the UVIVF capture happened to give almost identical result:

#15

2018-05-08-0250E-7251-subtracted-by-background-capture-md.jpg.20a50802a18da330b9bce1b8c9860898.jpg

Same capture as above edited. (But this adjustments only compensates for light from the night sky, not possible light leakage or UV leakage in filters on UV LED source and lens respectively).

 

 

Finally here is the correction of the original capture by increasing color temperature to about 8000 K. Again brighter tones get much warmer, compared to the effect on the darker tones:

#16


2018-05-08-0250E-7251-WB-colortemp.jpg.b44e10d29809d298811206402baffc5c.jpg

 

Same capture as above edited.

 

Some thoughts:

 

The corrections by subtractions are tricky due to the possibility of variable reflectance of the object of interest (or even different parts of it) compared to the reference patch. However if adjustments are small may be this can be acceptable (the situation of the outdoor test is not typical but rather extreme). The ideal calibration target should likely be gray rather than bright white while still not showing any fluorescence. Unfortunately the UVIVF signature of the gray patches on the Colorchecker shows brighter UVIVF emittance than the small white patches. There is also the possibility to do the correction "by feeling" but then standardized results cannot be obtained. So we are looking for a suitable (preferably cheap!) grey non-emitting standard.

 

The expensive commercial fluorescence WB targets tested here by Andrea,

https://www.ultraviol...iltered-uv-led/

seem to result in the "grey" color on the cold side in UVIVF when recorded with daylight white balance compared to a daylight lit grey object. This confuses me as a color should be the same for the sensor whether it is emitted in the dark with proper filtering or reflected off a gray object in daylight. Is there some inherent coldness in fluorescence that has caused the invention of the fluorescent WB standard targets that are doing an ad hoc. compensation for this? If so we are not seeing the "correct" daylight referred colors but UVIVF colors standardized so that they look good to the eye when using preset white balance on the targets.

 

May be even mild initial correction by light subtraction could be combined with WB/color temperature correction?

 

It would be nice to see experiments from others here with other targets and better scenes closer to a practical UVIVF capture situation and help with searching for grey non-fluorescent targets. Sometimes one can be surprised: when I tested the plastic tray that came with some frozen food and that I used for background in the indoor scene above I was sure it was going to be very bright in UVIVF, but it is one of the darkest backgrounds I have encountered so far. Personally I should look for a better UV cut filter for my lens than the L39. (It seems that the material of my polycarbonate UV safety glasses would have been perfect if the durability and optical properties had been better...)

 

Edit: Numbered images for easier referencing.

Edit2: Included more detailed technical info below each capture.

2018-05-08-0250E-7251-WB-graypoint.jpg

Link to comment
Andy Perrin
This is very interesting. I think you must be right about white balance by multiplying being a poor solution. The Baader UVIR cut is more reasonable priced than their U filter, and it has a pretty sharp cutoff at 400nm if I remember correctly. I do think you should be using more than 8bits per channel for this procedure. Rounding error due to subtracting will be extreme with only 256 possible values.
Link to comment
Thanks for the comment and filter tip Andy. I sure agree about working with more than 8 bits per channel. I see now why you made that comment, because of the numbers displayed on the histogram. I was performing these adjustments on 14-bit Raw in Capture NX2, and while we do not know the internals, it is highly unlikely that it works internally with less than 16 bits per channel. (It can open- and output 16 bit TIFF). The numbers are likely just a display thing, instead of showing a 0-100% scale as in the latest GIMP and RawTherapee.
Link to comment
Andy Perrin
I see, although such uncertainty is why I prefer to use MATLAB for this kind of thing! Then I know for sure.
Link to comment

Yes working on numbers are sure nice and subtraction is then just what it is, except for possible issues with working in gamma converted vs linear space, depending on what is used as starting point. (I do not pretend to know how to do it, although I am an occasional MatLab/Octave user.) Andrea sort of predicted your comment "in the frozen outdoor thread" :

" ((And I suspect that Andy P, when reading this, will also know how to accomplish this kind of subtraction in MatLab !!)) :)

Link to comment
Just a further thought: If one were doing this correction for background light pollution in MathLab, one could (preferably in linear color space) for each channel try to subtract the signal of the white non-emitting reference patch multiplied with a channel specific reflectance coefficient derived from a visual capture at each corresponding data point? I am not sure how this would work with wavelengths close to UV though as the visual visual light capture might not provide realistic reflectance coefficients?
Link to comment
Andy Perrin
I don’t know how you could find ANY reflection coefficients since you need information about both the incoming and reflected light to find those.
Link to comment

Øivind, thank you for your experiment and the write-up.

 

Reading with interest.

 

And I will come back later with a few more comments of my own after I have thought through some of it and can run a couple of tests myself.

 

*****

Before attending to the primary part of the experiment, let me put the following up for discussion (at some point).

 

Color Temperature: This is something I can't figure out for shooting in the dark. I'm not sure why Daylight would be a proper color temperature for visible fluorescence photographed in the dark? As sunlight wanes, the color temperature setting increases in our cameras. Taking that to its logical (???) extreme, we would arrive at somewhere between 10000 - 15000K for "shooting in the dark" depending on camera being used and converter being used.

I do realize that I might not have correctly phrased this thought.

 

 

 

So when Øivind writes the following, I think my answer is --> Daylight WB is not the correct temperature (K) in which to shoot in the dark. Daylight WB only works when shooting in daylight (sunlight). Later tests showed that the maximum Nikon setting of 10000K worked better for the fluor targets. (Although still not perfect.)

The expensive commercial fluorescence WB targets tested here by Andrea,

http://www.ultraviol...iltered-uv-led/

seems to result in the "grey" color on the cold side in UVIVF when recorded with daylight white balance compared to a daylight lit grey object. This confuses me as a color should be the same for the sensor whether it is emitted in the dark with proper filtering or reflected off a gray object in daylight.

Link to comment

I don't do much fluorescence photography myself, so forgive ignorance if it be too egregious....

 

Ideally, a fluorescence image records only light emitted by the subject matter. So if I understand correctly, a conventional reflectance target is meaningless as a standard in this context. If there is no extraneous background in the image, the only way I can see to standardize this is to balance it with respect to some sort of standard phosphor target placed in the image, whose emission under the circumstances is agreed upon as "white" (exactly what tint is produced will depend on the profile of the exciting light.) I assume that the commercial targets referred to above are some sort of standard phosphor. The "daylight" fudge might produce similar results if the output of the standard phosphor under the UV source in use has approximately the same temperature-and-tint as sunlight, as recorded by the sensor. But white balance in a fluorescence image seems to have a degree of arbitrariness over and above what we encounter in reflectance images--we lack any unique, natural definition of what "white" means.

 

Add contamination and it gets even more murky. I think that for subtraction to work, one would need a whole calibrated greyscale of standard targets that are known not to fluoresce. Furthermore, they would need to be positioned in such manner that light emitted by the subject itself does not strike them, but any stray light in the scene does. It might be better to take two exposures, one with the subject matter and one with the reflectance greyscale, so this sort of confounding problem does not occur. Any attempt to do this with only a single reference target is likely to run into problems.

Link to comment

I don’t know how you could find ANY reflection coefficients since you need information about both the incoming and reflected light to find those.

 

Andy, I agree it is problematic, but I think one could calculate a reflectance value that is relative to the white non-emitting target, also located in visual image (It will not be true reflectance by the definition). After all what we want to compensate is how different parts of the subject of interest deviates in reflecting light pollution relative to the white target used to measure background light pollution in the UVIVF capture. However it would require that the daylight reference image used to determine spacial relative reflectance to be perfectly aligned to the UVIVF capture, so in practice it might be difficult to perform.

Link to comment

...

Color Temperature: This is something I can't figure out for shooting in the dark. I'm not sure why Daylight would be a proper color temperature for visible fluorescence photographed in the dark? As sunlight wanes, the color temperature setting increases in our cameras. Taking that to its logical (???) extreme, we would arrive at somewhere between 10000 - 15000K for "shooting in the dark" depending on camera being used and converter being used.

I do realize that I might not have correctly phrased this thought.

...

 

Andrea, the change in color temperature as we approach evening has little to do with the light intensity as such (thus darkness) but how much atmosphere the sun passes though on its way to us causing warming of the light source, and once the sun is down it might at least for a while reverse in the "blue hour". If the moon comes up that is actually again a fairly warm colored light. The white balance is just something we use to compensate for color temperature and tint of the light source that is then reflected off our subjects, The color emitted from our objects of interest will not change due to the darkness. In astrophotography the colors are adjusted as for a light adapted eye, that will provide the right colors emitted from white stars like our sun, and make the red and blue stars appear as we would see them if our eye is sensitive enough. In the DPreview astro forum Roger Clark (and his informative web pages) recommends the experiment to bring a color chart along under a maximally dark sky in a dark zone without moonlight after dark adapting the eye for more than an hour and see how colors appear. The ones that can be distinguished he writes that they are very close to what the eye would see in daylight. As a physiologist I am a bit more skeptical (never tried the experiment), as our eyes adapt very much to the overall background color hitting the area of retina containing the cones, so absolutes can get pretty shaky (just think of lateral inhibition clearly visible when we watch a color chart with colors next to each other or a simple grayscale).

 

However from a colorimetric point of view one would have to adhere to some kind of standard, and a color sensitivity of our capture and processing chain close to that of how a color chart would look like with a daylight light source and daylight white balance would give the closest realism for viewing emitted colors with a daylight adapted eye. Most high quality computer monitors (also emitted light) when properly calibrated today will also be set so that a camera set to daylight WB will record fairly realistic colors. Would one use a different WB if recording or viewing the computer monitor in complete darkness?

 

The nice tests you performed with good filtering on the emitting standard targets (that apparently are used extensively in forensic sciences) indicates that they only work with a preset WB, while automatic WB and daylight WB give pretty similar results on the cold side. It is possible that the camera's logic decides that it cannot find a reasonable grey color in the image, and cannot find colors that it would make any sense to average out, so it defaults to something very close to daylight WB.

 

But again we are back to how colors look like in UVIVF and what is causing it. If not possible to use for a mathematical subtraction of the light pollution, it seems to me that the inclusion of a white non-emitting standard along with the emitting standard forensic targets (but not so close that they are influenced by the emitted light) could be used to evaluate the maximal effect that light leakage and other light pollution of the setup could have. If a white non-emitting target becomes pretty much black, then we know that the coldness of those emitting forensic standards with daylight WB is not light pollution (and that the forensic standards are designed to try warm the emitted colors to make them more pleasing to the eye on a standard monitor). It could also be used in a similar way for practical scenes.

Link to comment

I don't do much fluorescence photography myself, so forgive ignorance if it be too egregious....

 

Ideally, a fluorescence image records only light emitted by the subject matter. So if I understand correctly, a conventional reflectance target is meaningless as a standard in this context. If there is no extraneous background in the image, the only way I can see to standardize this is to balance it with respect to some sort of standard phosphor target placed in the image, whose emission under the circumstances is agreed upon as "white" (exactly what tint is produced will depend on the profile of the exciting light.) I assume that the commercial targets referred to above are some sort of standard phosphor. The "daylight" fudge might produce similar results if the output of the standard phosphor under the UV source in use has approximately the same temperature-and-tint as sunlight, as recorded by the sensor. But white balance in a fluorescence image seems to have a degree of arbitrariness over and above what we encounter in reflectance images--we lack any unique, natural definition of what "white" means.

 

Add contamination and it gets even more murky. I think that for subtraction to work, one would need a whole calibrated greyscale of standard targets that are known not to fluoresce. Furthermore, they would need to be positioned in such manner that light emitted by the subject itself does not strike them, but any stray light in the scene does. It might be better to take two exposures, one with the subject matter and one with the reflectance greyscale, so this sort of confounding problem does not occur. Any attempt to do this with only a single reference target is likely to run into problems.

 

OlDoinyo, thanks for commenting. If you look in the stickies link, Andrea did some tests on some emitting targets, which seem to be biased towards warming the colors beyond what a daylight adapted eye would see (say in the hypothetical situation that we could have amplified the light somehow to view in a dimly lit room at 5000K background illumination). So with daylight color balance on the camera, they look cold.

 

I am not sure a greyscale non-emitting target would be practical in performing differentiated subtraction in the UVIVF capture, but it could be useful to evaluate the effect of for instance light leakage in the UV source on different parts of our object of interest with different daylight reflectance. It could also be used for selecting a proper average subtraction level in cases of moderate light pollution. I suspect in most cases we will actually see very small or almost negligible effect of light leakage if we pay attention to filtration and darkness. Light emitted from backgrounds and surroundings and in turn reflected from our target may be a different matter. Such a grayscale non-emitting standard could be very useful for evaluating and planning/modifying the scene to improve in this respect. I think it is pretty common that we see colors in our UVIVF captures where we wonder if they are emitted or reflected. (Examples are the UVIVF appearance of neutral or green colors with moderate intensity of parts of plants that are green in VIS). The inclusion of a non-emitting reflecting standard could provide the answer.

Edit: there was also a good discussion of this in the "frozen outdoors" thread.

Link to comment

Question about the white non-emitting standard: does it reflect UV or absorb UV? That is, I'm asking which definition of "white" are you using? Reflect all of UV, R, G and B? Or reflect just R, G and B?

 

The UVI targets which fluoresce white are calibrated for a 360 - 370 nm range and assume that a Wratten 2E longpass is used on the camera lens to block UV/violet and some blue together with a Peca 918 to block IR. The assumption is also that the camera is stock. If the camera is modified, then an additional BG-38 is recommended. So these targets define a "white" for UVIVF only in a narrow range of UV illumination and slightly restrict what the camera records to approximately 420 nm onward to approx 690 nm (700.?)

 

When originally testing the targets, I attempted to find an in-camera white balance setting which people without fluor targets could use in UVIVF with their Nichia 365 UV-LED torches. That turned out to be a high K setting for my particular camera. Whence my backwards reasoning above. :)

 

 

Edit: added closing range 690.

Link to comment

Tricky question, especially since some of the potential light leakage is in the transition. :) We would want the standard to appear white/grey in VIS only I think, not when UV is added. Whether UV is reflected or absorbed, would not a proper UV block filter take care any UV reflected back from the white non-emitting standard? With the stated parameters filtering somewhat into blue it should not be necessary to reflect UV. Also UV sensitivity of the sensor is low compared to visual light.

 

From stickie's experiment 3 I got the impression that a grey-point compensation was used, but may be I did not read it well enough and a kelvin value was used? (The greypoint compensation could also affect tint). "Made against the UV-Grey Target under the 365nm UV-Led torch in darkness. A new Preset WB was made for each torch and lens filtration combination."

 

Would you by any chance be able to make the NEF files for that experiment available? There is not much to play with in the JPG's.

 

I looked for a source of non-gelatineous Wratten 2E (I.e a 52mm screw in filter) but did not find it so far (looked at B&H). Any advice?

Link to comment

Wratten 2E as a longpass cuts in at 415nm. (But we need to find a chart for it to determine how "cut-in" is defined.) So the nearest longpass I know of is Schott GG 420nm, also a pale yellow filter which you should be able to get from Uviroptics. (We need to appropriate determine thickness.)

 

HOWEVER --->>>> GG 420nm fluoresces. So it needs to be used in such a way that its fluor does not affect the photo.

 

I note that there are filter adapters which adapt square filters to a round holder. So use of a Wratten might be feasible.

 

Going now to look for Wratten chart and that filter holder.

Link to comment

Square filter holder: https://www.edmundop...Imaging-Lenses/

That is rather expensive!!

 

Here are charts for the Wratten 2E and Peca 918.

http://dev.imagescie...era-filter.html

I note that when combined they provide a transmission fairly similar to the Baader UV/IR cut in the sense of having steep shoulders between 410-420 nm and between 680-700 nm. The Buvircut has its left shoulder further left between 400-410 approximately.

 

 

So If you combine a GG420 with the Peca 918, then the left shoulder moves a bit to the right of the Wratten 2E + Peca 918. But still looks quite similar.

 

To summarize:

  • Wratten 2E + Peca 918
  • GG 420 + Peca 918
  • Baader UV/IR-Cut

And note that all three of these filtration options will have a tiny bit of transmission past 700. So UVI recommended the BG 38 as an addition when using a converted camera. But I think I would use some S8612 (2mm) or BG 39 (2mm) because they taper off before 700 nm and the BG 38 (2mm) does not.

 

I am not sure why UVI suggests "adding" the BG 38? Once you add a BG 38 or 39 or 40 to a Wratten 2E + Peca 918 stack, then the Peca no longer contributes anything. So may I have misread that and UVI is suggesting using a Wratten 2E + BG 38 instead of the Wratten/Peca stack?

 

Here is what the curve would look like for GG 420 + BG 39. The steep shoulder on the right is replaced by a more normal curve for the reds.

 

This is a linear transmission chart with OD4 suppression starting at 700nm.

GG420x200_BG39x200.jpg

Link to comment

Øivind: From stickie's experiment 3 I got the impression that a grey-point compensation was used, but may be I did not read it well enough and a kelvin value was used? (The greypoint compensation could also affect tint). "Made against the UV-Grey Target under the 365nm UV-Led torch in darkness. A new Preset WB was made for each torch and lens filtration combination."

 

In experiment 3, I wanted to test the outcomes for 3 different types of filtration on my 365nm UV-LED torch.

For each of the 3 filtration choices, I made two series of photos:

first with an in-camera WB set against the UV-Grey card and second with the Daylight WB setting on the camera.

So that was 6 sets of photos.

When I moved to the next filter choice, I had to re-shoot the WB preset against the UV-Grey card.

 

The first outcome of the experiment was that using an in-camera WB preset against the UV-Grey card for a particular filter choice proved to be more accurate than using Daylight WB. By "more accurate", I mean that in the photographs the white/red/green/blue patches on the Target-UV card were white/red/green/blue. No edits were made on the raw files. The results were analyzed straight-out-of-camera.

 

The second outcome of the experiment was that the different filtration on the 365 UV-Led torch seemed to make no difference in what the photos showed.

 

It was only later in Experiment 4 that I attempted to correlate the "more accurate" photos to a particular K temperature dialed into the camera as a white balance setting. It proved unsuccessful because a Nikon can only go to 10000K. Analyzing the photos in Photo Ninja showed that the accurate photos had a K between 13000-15000. (In Photo Ninja when you use the white balance dropper, you can see the new Temperature and Tint settings for the WB step.)

 

I hope that clears it up? If not, I'll try again. :D


 

In those experiments I think I did not make it clear that one might not get the same accurate results if using a UV illumination different from 360-370 nm.

Link to comment

Thanks so much for the information Andrea. With a modern non-modified body that already as a fairly strong UV-IR cut filter, I am not sure that there is a point of filtering on the IR side? Consider my spoon test with my IR only body in image #8 (I have now numbered them), IR from the filtered UV-light source does not seem to be anything to worry about. There is only a hardly visible spec in spite 30s exposure; note that I am using the less IR leaky 2mm ZWB1 on the front of the light and it has in addition a dark front glass (possibly a ZWB2). There is of course the potential for UVIIF contamination, but again one would not expect any significant amount of that to get through the standard UVIR cut filter on the sensor. So with the $190 price tag :blink: the Peca 918 seems a bit of an overkill. (A wideband body would of course be a different matter.)

 

Do we have any transmission curves for any newer factory standard on-sensor UV-IR cut filters anywhere? I thought I had seen it somewhere here at some point? I found some at the Kolari Vision site, https://kolarivision...r-transmission/ ,but for very old bodies. With members here ripping those filters out of the body combined with members with acess to spectrometers, someone would have tested them?

 

Fluorescence of the Schott GG 420nm seems possibly problematic to me (as do that of potentially occurring in the lenses, but not with a UV cut filter on the front...).

 

There is also the concern of cutting too much of the visible spectrum when doubling up filtering. Curiously the Baader UV-IR cut filter is depicted with pink hue, https://www.baader-p...--l-filter.html, seems quite opposite of what I would have expected when cutting at the short and long end of the spectrum and even slightly into the blue?

 

The filter sizes seem rather odd, for my lens 50.4 and 50.8 (2") could possibly work without vignetting with suitable adapters to 52mm thread, I have never heard of step rings to those sizes? Any tips as to how to mount them?

 

(Edited in parallel with Andrea's last message)

Link to comment

Pertaining to post #17.

If you want less 700nm+ then just use thicker BG39 or S8612, those have identical red/IR curves and suppression at the same thickness.

BG39 has less UV and less blue transmission than S8612 at the same thickness, so I might expect BG39 to look slightly less blue than S8612 at the same thickness.

Bluer < S8612 - BG39 - BG40 - BG38 > Redder (all at 2mm).

post-87-0-83763300-1526162642.jpg

 

post-87-0-99594000-1526162662.jpg

Link to comment

Thanks for the clarification Andrea. The sentence:

"The second outcome of the experiment was that the different filtration on the 365 UV-Led torch seemed to make no difference in what the photos showed."

is enlightening, perhaps we are worrying too much about light leakage (as long is we have some kind of filtering on the light) and should worry more about influence of reflected emittance from the surroundings?"

 

I wonder how those standards were established in the first place. It seems pretty clear from your results it is not because of a tendency of blue light leakage in UVIR captures, but the perception that they on average become too cold and need some warming. Did you notice any changes in the tint values resulting from the custom WB?

 

Would you be willing to share one of those NEF files if it is not too much trouble to dig one up? it could be fun to get some first hand experience in playing with one of them.

Link to comment

The filter sizes seem rather odd, for my lens 50.4 and 50.8 (2") could possibly work without vignetting with suitable adapters to 52mm thread, I have never heard of step rings to those sizes? Any tips as to how to mount them?

 

My 2-inch Baader filters have all been pre-mounted in 48mm rings by the maker. I add step rings as needed and have experienced some minor vignetting on some lenses Seems ok on the UV-Nikkor.


 

BG39 to look slightly less blue than S8612 at the same thickness.

If you have removed the original internal filtration from your camera and then want proper visible color, a color profile made on each lens+filter combo seems to be a necessity -- in addition to the usual white balance adjustment. I've seen some white balanced BG photos here on UVP which look "off" in their color because I suspect no color profiling has been done?

Link to comment

"My 2-inch Baader filters have all been pre-mounted in 48mm rings by the maker. I add step rings as needed and have experienced some minor vignetting on some lenses Seems ok on the UV-Nikkor".

Thanks Andrea. I see, I thought those were thread sizes not glass sizes. My 105 f/4 micro should be similar to the UV nikkor - less sure about the 55mm f/3.5 micro. Some simple experimenting should tell.

Link to comment

perhaps we are worrying too much about light leakage (as long is we have some kind of filtering on the light) and should worry more about influence of reflected emittance from the surroundings?

 

I think that the experiments tend to support that conclusion for the Nichia 365nm torches. As mentioned before (where? elsewhere somewhere...) if using a 375 or 385 or 390 nm torch, then you are going to get a lot of spillover into the visible and cause visible reflection "contamination". So filtration for those kinds of UV illumination might be more important?


 

But back to your method....Above you mentioned that astro photographers try to adjust their photos to a daylight white balance as though the stars were being seen with a daylight adapted eye. And I wanted to say that I understand why that is done. And it makes sense in that it provides a standardized astro photo output so that judgements can be made relative to our Sun.

 

So why is the adjustment to a daylight WB and a daylight adapted eye not working for these fluorescent targets I was using? I don't know!

 

The goal for these fluor targets is that the camera which has a preset made against the UV-Grey card under 365nm UV will record the fluorescent grey/white/red/blue/green Target-UV patches as exactly that -- as grey/white/red/blue/green patches. But what happens when I look at those color patches in the dark under the 365 UV torch? I'm going to go find out. I'll go into the dark closet with my light adapted eyes and quickly turn the torch on those colors, observe them and then come back and tell you what I'm seeing.

 

BRB

Link to comment

Looking forward to the results. :)

 

Perhaps add another control experiment with say 10-15 minutes in the dark to compare impressions?

Link to comment

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...