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

[Andrea B.]

And I've just noted that my eyes are currently adapted to rather a mixture of late day daylight and tungsten "warm" LED light and TV light and various other ambient bits-and-pieces of light.

 

Here's what I saw on the fluor Target-UV patch card: white, grey, grey, red, green, blue.

(For the record, that was with No Goggles.)

I asked the SigOth to look quickly into the dark at this target illuminated with 365 nm UV and tell me what colours he saw.

The same.

I asked did he see any color cast or tints? Did the whites or greys look a bit blue or a bit pink or whatever?

He said No.

 

Now that is admittedly not a rigorous experiment. But it does tell me that this particular set of fluorescent standards tuned to the 360-370 nm UV inducing range seems to match my light adapted eyesight.

 

If that is so, then why does the Daylight setting on the stock D810 camera not produce the same white, grey, grey, red, green, blue?

I don't know. 'Tis a mystery !!

 

As a side note: I just observed that some of the Target-UV patches are very very slightly phosphorescent and take a couple of minutes to fade out.

 

So this is enough straying Off the Topic of subtraction of contaminating light. I'm still thinking through how I might try an experiment using some of the reflective standards I have. If the emitted fluorescence shines on the standard and we subtract it, then we are removing some of the fluorescence. So the set-up involves contaminating the standard while preventing emitted light from shining on to it. Did I get that right? Thus, I think if I set up the white standard in front of the Target-UV patch card, then the emitted fluor from the patches would not shine on the standard. I'll use one of my little micro LED lights as contamination. I may not have enough hands for all this.

[otoien]

Thanks for reporting the test results. While there is scotopic light adaptation over time (which is the effect eliminated in this experiment) our brains are pretty quick to adapt to differences in background illumination. If we go from a room with daylight light source to one with warm LED or incandescent bulbs we will still not say that that red object now looks orange (although we are aware that the room has warmer light). In a dark room it is a question how things are interpreted with no surrounding reference and very strong contrasts. (Also remember that blue - or was it white - dress that went viral?). It is possible that when our brain detects something close to gray or white that it will interpret that as gray or white even if has a tint. It could have been interesting to do the same experiment with a very low intensity light box in the room with a color temperature that the camera interprets as white/gray with daylight WB, only exposing small patches that are shielded from the fluorescent reference so that they are not influenced. How would the eye interpret that? I would predict that the fluorescent gray patches now look colder. But who knows, may be it will even differ between individuals which of the patches, visual light or the fluorescent gray/white that will be interpreted as neutral...

 

One could also try to display the images with daylight WB of you fluorscent patches with a black background on a calibrated monitor in a completely dark room. Then bring up beside it the ones with preset WB. Which will be interpreted as gray in the two situations?

[otoien]

To answer my own question, I found another Kolari Vision page with curves for many factory on-sensor filters. From that it seems that the cutoff is already at 420nm,

https://kolarivision...rum-conversion/

A screen shot:

 

So with a 365nm LED light source it seems to be overkill to add a UV-IR cut filter on the lens in addition to an unmodified body's internal UV-IR cut filter unless the body is quite old. The blue curves for the UV cutoff in the figure looks very much like the Baader Panetarium UV-IR filters cutoff curve. (Of course a body modified to wide spectrum does require filtering and makes the suggestions above useful.)

 

Edit: For the record of the thread added a link to recent tests in this forum: https://www.ultravioletphotography.com/content/index.php/topic/2758-eos-5ds-r-sensor-filter-transmission-measurement/

 

If the lens shows internal fluorescence that is another case for filtering, https://www.ultravioletphotography.com/content/index.php/topic/2289-lens-glass-fluorescence. Compensating that would be a good case for subtraction as suggested by OlDoinyo in the linked thread. Even a standard not so strong L39 UV filter on the lens might help considerably. Here is my 365 nm LED light shining at my 105mm f/4 micro, at ca 10 cm distance without filter on that lens:

#1

 

NIKON D5100 non-modified, 55mm f/3.5 micro @ f/5.6, 8s, ISO 100 and 365nm UV LED light as noted at the start of the thread. No filter on the 55mm. The depicted 105mm has no filter attached.

 

 

L39 on the 105mm f/4 reduces internal fluorescence considerably, also note that the red line at the edge is gone:

#2

 

NIKON D5100 non-modified, 55mm f/3.5 micro @ f/5.6, 8s, ISO 100 and 365nm UV LED light as noted at the start of the thread. No filter on the 55mm. The depicted 105mm has a Nikon L39 filter attached.

(Both captured under kitchen sink before it got dark, so not ideal with respect to background illumination).

 

The intensity would of course be much lower in a practical situations compared to shining the light into the lens. (As known, shining flashlights into lenses is a good way to go insane .)

[Andrea B.]

(1) I'm confused by the Kolari comment. If the red line represents their UV/IR blocker, why is it not blocking UV?

 

(2) From looking at all those Original Lowpass Filter curves, I am fairly well convinced that a current digital camera's internal filtration is close to what we see with a 420nm Longpass stacked with some type of BG glass (I just showed a similar curve in a GG420+BG38 chart somewhere.) Although if longpass/BG filtration is indeed used for internal filtration, I wonder how they deal with the GG fluor? Perhaps with some coating?

 

(3) So with a 365nm LED light source it seems to be overkill to add a UV-IR cut filter on the lens in addition to an unmodified body's internal UV-IR cut filter unless the body is quite old.

I tend to agree. But if there is any IR fluorescence just past 700 nm, you might pick that up. How much any fluor around that 700nm "border" should bother us is up to an individual to decide. After all, we are not making scientific studies with our UVI fluor photos. If we were, then we would be using very narrowband filtration on lens and on light.

 

(4) Would everyone agree that the important thing is to provide a few details (lighting, filters or not, camera mod or not) about one's fluorescent photograph so that everyone understands what they are seeing?? I have tried to encourage that here anyway. :)

[otoien]

...

(1) I'm confused by the Kolari comment. If the red line represents their UV/IR blocker, why is it not blocking UV?

...

I did not want to focus on Kolari's own filtration and the red line (their comment just included because of the 420nm statement) as I was also confused by the marketing as a UV cut filter. For instance almost no UV would be removed from a 365nm LED source . Added Edit: However that is an old filter. Below that in the link provided, there is a newer UV cut filter showing proper transmission properties.

 

I certainly agree to #4 that we need to provide as much information about UVIR captures as possible. Edit: To follow that up I tried to enhance the technical info below each capture in my posts above, not only in the main text as that sometimes can be missed.

 

At 700nm, only 2-4% would be transmitted, and in my experience, the UVIIF requires longer exposures than UVIVF, in the examples above nearly 4x the exposure and still dark. So yes, I think we can live with one or two percent IR contamination of our UVIIF captures :) . (Our daily non-mod body visual captures have similar levels of IR contamination - no one complains much about that).

[Andy Perrin]

(Our daily non-mod body visual captures have similar levels of IR contamination - no one complains much about that).

Doesn't it really depend on what fraction of the captured light is Vis relative to IR? It's not just the filter that matters. While in sunshine it's roughly the same fraction of visible and IR (55-45% or something like that), I think it would be subject-dependent in fluorescence.

[otoien]

Yes of course the fraction can vary, both in a UVIIF/UVIVF and a daylight situation. (Added clarification: I was referring to visual daylight scenes captured with a non-modified body in that sentence, and in that case the dominant IR contamination would likely be the much stronger reflected IR.)

 

A Nikongear post recently referred to a web page where someone thought they got good IR response from an unmodified D850. The exposures with a IR pass filter attached to the lens were 30 seconds where an IR modified body on the same outdoor scene would have used 1/100 sec or shorter.... If the UVIIF image above (#8) is underexposed with 5EV in post (a simulation of transmitting ca 3% of the whole captured IR spectrum, not just that around 720nm) it is almost black, I can just barely make out the egg. Underexpose with another 2EV in post (to compensate the longer exposure) and one would will not be able to make it out, the displayed 8-bit value ends up at 2 on the egg.

 

If someone happen to have an IR pass filter, a recent non-modified body and a little spare time, why not try it on a subject that is known to provide a particularly good UVIIF response relative to UVIVF and see what one gets (using a regular UV pass filtered 365nm LED flashlight and without using crazy ISOs of course...). :)

[Andrea B.]

OK, I can do that one.

Let me make sure I have the parameters you want:

• D850-stock
  
• ISO between 100-400
  
• Convoy 365 UV-LED illumination
  
• IR-Pass filtration: I think I will use my 3 IR-bandpass filters from MaxMax.
  

However, I'm not sure what you were suggesting for a subject?

You wanted something which is both Vis fluorescent and IR fluorescent under 365nm UV illumination?

[otoien]

"You wanted something which is both Vis fluorescent and IR fluorescent under 365nm UV illumination? "

 

Good IR fluorescence was the idea, so we do not care so much about a good UVIVF response. But it would be a good idea to keep track of how much exposure one would have normally needed for a UVIVF capture of the same subject to figure out the relative UVIIF contamination of the UVIVF capture. So may be a test with and without the IR pass filter?

[Andrea B.]

I ran a preliminary experiment. Here is the data. I'll post some photos tomorrow. (Getting late here.)

 

Cameras: D850-stock and D610-broadband

Lens: UV-Nikkor 105/4.5

Settings: All photos f/4.5 @ ISO-400.

Lighting: Two filtered Convoy torches were shined on the Target-UV fluorescent target in the dark.

 

Both cameras were white balanced against the UV-Grey fluorescent card under the double Convoy UV illumination.

 

D850-stock UV-induced Visible Fluorescence

• Baader UV/IR-Cut: 1/80 second
  

D850-stock UV-induced Infrared Fluorescence

• Schott RG9 (2mm): 30 seconds, photo somewhat dark compared to subsequent photos on the D610.
  
• Schott RG715 (2mm): 30 seconds, slightly brighter than with RG9.
  
• Schott RG780 (2mm): 30 seconds, got nothing, all black.
  

D610-mod UV-induced Visible Fluorescence

• Baader UV/IR Cut: 1/50 second
  

D610-mod UV-induced Infrared Fluorescence

• Schott RG9 (2mm): 1/1.6 second
  
• Schott RG715 (2mm): 1/4 second
  
• Schott RG780 (2mm): 3 seconds
  

So if the question is whether the stock D850 is good for UVI Infrared Fluorescence, then the answer is definitely No in my opinion. And remember I was shooting at the UV-Nik's widest aperture, f/4.5. That's fine for flat cards and targets, but everyone would probably want to stop down a bit more if shooting a more dimensional subject. So that neccesitates either longer exposures or higher ISOs or both, which would become rather painful, methinks. You really are forcing IR light through a D850 system. Why fight that internal filtration?

 

(( I did not use my IR bandpass filters from MaxMax because I couldn't find a step-ring for them!))

[otoien]

Thanks for reporting and taking the time to test Andrea.

Those convoys must be quite strong, or that target shows really strong fluorescence with such short UVIVF exposures, even when considering the aperture and ISO. Not so surprising about the non-modified D850 response. Looking forward to the captures. Does not seem that the D850 UVIVF captures would have any significant IR contamination, considering the differences in EV values.

[Andrea B.]

The problem area is always around 700 nm. These borders between UV & Visible and Visible & IR are a little bit arbitrary. And it is between 680-730 nm (more or less) that we have problems with filters "leaking" light which can be considered either high red or low IR. (my terms). You will see in the photos I'm about to post that the D850 can record something there.

[Andrea B.]

Notes

The NEFs were converted to TIFs in Photo Mechanic.

The labels were applied to the TIFs in PS Elements 11.

The TIFs were downsized and saved as JPGs in Photo Mechanic.

No other edits were made.

White balance was set in-camera against the fluorescent UV-Grey card.

 

There is an approved methodology for determining the proper exposure for the Target-UV when making a UVI visible fluorescence photo. However, I eye-balled it this time. You'll see why this was a bad idea when you look at the first two vis fluor photos from the D610. Thus we learn the hard way to follow directions.

 

I have not removed the cat hair contamination.

Kitty Mocha literally walked through the set swishing her fluffy tail. Sorry 'bout that.

 

Test Photos from D610-mod

 

There is some color contamination at the bottom of the first UVI vis fluor photo which I attribute to uneven illumination with the Convoys. The next UVI vis fluor photo does not show this because I aimed the torches more carefully and made a longer exposure.

 

 

However this version of the vis fluor, while not color contaminated, is over-exposed. As I said up there, Andrea should follow the directions next time. (embarrassed giggle)

 

 

Next we learn that cat hair is not IR fluorescent under UV illumination. OK then.

White balance remains at the Grey-Target in-camera WB preset made for the vis fluor photos.

 

 

No particular comment to make here.

 

 

Note the very slight shift in focus when we jumped from 715 to 780nm. I did not re-focus.

 

 

Test Photos from D850-stock

 

In the preliminary test with the D850-stock, I could not capture any UV-induced IR fluorescence at the base ISO-64 when using a 30 second exposure time. I had to go to ISO-400 to begin to see the IR well enough to make the test photos. So to fairly compare, all photos for both cameras were made at ISO-400.

 

 

Nothing surprising from the UV-induced visible fluorescence photo from the stock D850. It's a little bit underexposed but we know why (Andrea was careless!).

 

 

I stopped at 30 seconds to make the point that IR, whether reflected or emitted, is difficult to shoot with the stock D850 which has strong internal filtration. This shot could have been properly exposed by going beyond 30 secs or by boosting ISO, but we would get noise.

 

 

The nature of the RG715 IR-longpass makes this next exposure a little easier than with the preceding RG9. And with a 30 second interval I could ensure even light painting. "-) The equivalent D610-mod photo required 1/4 sec. So from 1/4 sec to 30 sec is 7 stops. I was definitely fighting the internal filtration to get this shot.

 

 

Got nothing at all at 30 seconds with the RG780. The internal filtration cut this IR out quite well. If you push this file hard in a converter (3+ stops) you can find the Target-UV in there, but the noise is hideous. I'm not sure how many actual shooting stops you would need past 30 seconds to raise any kind of image. But why bother?

[otoien]

Thanks for the test Andrea, very nice to see. So with stock D850 we have >10 stops difference in exposure between UVIVF and UVIIF, so again, safe to assume that the UVIVF exposures were not influenced by the IR portion in any way we would care about. What is the background that appears to be neutral grey as the standard target (or is it part of the target)? BTW, it almost seems that the D850 UVIVF exposures gives cleaner colors on your target?

[Andrea B.]

Øivind, I think I need to repeat this test. I was using a Baader UV/IR cut filter on the UV-Nikkor. I want to run the test again with the D850 and the UV-Nikkor but without the UV/IR cut filter on the lens. The UV-Nikkor will then pass any emitted IR light and we can more accurately test whether the internal filtration of the D850 blocks it.

 

These photos were completely unedited so it is probably best not to make any judgements about the colors. Typically I would adjust histogram endpoints (black pt and white pt) and adjust for channel oversaturation (blown highlights). Also when converting files from a modified camera it is necessary to apply a color profile to compensate for the missing original filtration which typically has a slight color tint. That was not done here for the modified D610.

 

All that said, the Nikon D850 does have exceptionally good color, the best I've ever seen. :)

 

The background is the UV-Grey fluorescent target. It is rectangular and larger than the small strip of the Target-UV.

[otoien]

Thanks for the explanation, I see, so the D850 captures were double filtered, first by the Baader UVIR cut filter and then by the one on the sensor. Interesting that the test let though anything at all then, but of course what is getting though is in the transition around 720nm.

[Andrea B.]

I wanted to make the UV-Nikkor behave more like an "ordinary" lens and not easily pass the Infrared.

However, given that I was supposed to be testing the D850 sensor against IR, maybe that wasn't the best idea.

[otoien]

Now I got curios - how much less IR would for instance the 105 f/4 micro (an otherwise pretty close construction) pass relative to the unfiltered UV nikkor? I have never gotten the impression that IR was much attenuated by lenses or to a variable degree, but might not have paid attention. (I do notice that if I shoot through the windows of my cabin, it does appear attenuated/more red slightly attenuated/slightly warmer with preset WB for outside IR.)

[Andrea B.]

Which is the 105/4 micro? I have a Micro-Nikkor 105/2.8D. I think it hotspots in IR though.

[otoien]

Hotspotting is of course always a potential complication. Mine is the 105mm f/4 AIS micro, the 105mm f/2.8D woud be a more complicated construction with retrofocus mechanism. Would you also happen to have a 105mm f/2.5? The best I could do was a quick test to compare my 50mm f/2.8 Cassarit to AFS 50mm f/1.8 and AF 28-105mm /f3.5-4.5, all at 50mm f/8 and 1/50 sec ISO 100, no filter on lens, flat light in overcast weather, preset white balance. None of these lenses hotspot at the chosen aperture with my IR body. I also compared a capture through window panes with that of outdoor, all with my D40x IR-720nm (Lifepixel standard conversion). Exposures were evaluated by making adjustment in ViewNX2 until exposure looked similar.

 

Results:

 

Though window pane relative to outdoor: -0.6 EV, only slightly warmer.

 

AFS 50mm f/1.8 relative to 50mm f/2.8 Cassarit: -0.6 EV, only slightly warmer.

AF 28-105mm /f3.5-4.5 relative to 50mm f/2.8 Cassarit: -1.1 EV much warmer.

 

There was indeed some differences beyond what could have been expected due to inaccuracy of the aperture mechanism, from brightest to less bright dependent on complexity of the lens construction. (Perhaps the number of lens elements is more important than type of coating?)

 

My memory was incorrect with respect my window panes, much less effect than what I remembered (corrected above).

[Andy Perrin]

You can’t conclude anything from a test with one window. There are a wide variety of windows, many with IR blocking coatings. For instance my bedroom window cuts NIR by at least a stop with a 720nm filter and by several stops in the 1000nm range.

[otoien]

Yes of course, the glass itself could be different, single, double or triple panes, and as you point out IR reflective coatings. I did not pretend for the above to be anything but a case report to correct my initial statement about IR absorption in the plain single window panes of my cabin (for clarity added "my" to the last sentence of the previous post).

[Andy Perrin]

Oh, I see. I think I misunderstood the point of the test, then.

[otoien]

For the record of the thread I added a link to recent on sensor filter tests in this forum under reply #28.