Flowers with two false colours

[Andy Perrin]

Well, if all of this is so hard to interpret, why do we even bother with false colors in UV?

 

Normal camera uses three color channels to record and interpret 300nm wide band of visible light (400-700nm roughly).

 

In UV, we record a band of wavelength that is barely 50-60nm wide. Would it not be more logical to treat it all "together", as single "channel", by converting it to grey-scale, or, even better, using monochrome sensor for UV-photography?

The logic here is not quite right because color actually depends on photon energy, not wavelength at all. We only use wavelengths as a proxy. Using photon energy and replicating the rest of your argument,

320nm corresponds to 3.75 eV (Note: eV are electron Volts, a unit of energy, for non-Alex people following this)

400nm corresponds to 3.10 eV

700nm corresponds to 1.77 eV

 

If we go with the standard 7 colors for visible light (ROYGBIV), the average spacing is (3.10-1.77 eV)/(6 gaps between colors) = 0.22 eV between each color. So our UV band should have

(3.75-3.10 eV)/(0.22 eV) = 3 gaps, or 4 colors.

[Alex H]

Please note that I suggested to treat UV pictures as single "channel", not single color. Each channel of RGB sensor does cover the wavelength and photo energy responsible for more than one color in the standard seven color list, but as long as we use RGB monitors and RGB eyes, we have no way to show seven colors without using other channels and complex interpolation mechanisms in our brain. How do we do it with UV channel using RGB "environment" ?

 

Instead of using wavelength of ROYGBIV colors as proxy, and converting them to photon energy, please use wavelength sensitivity ranges of RGB channels as a proxy, and do the same.

[Andrea B.]

Well, if all of this is so hard to interpret, why do we even bother with false colors in UV?

 

(1) Bjørn and I wanted to standardize the appearance of the botanical UV-signatures across gear platforms for posting in our UV-Signatures section. There are two ways to do that. Convert to monochrome or use a white-click white balance tool in the converter to produce a false colour palette. We prefer the false colour.

 

(2) Some folks like false colours because they lend a more artistic appearance to some UV photographs.

 

(3) There have long been attempts to use false-colour in a UV photograph to determine the reflecting wavelength ranges. The question is: can it be done accurately and precisely? Well no, imho. But it is rather fascinating to investigate. And I've never been one to give up just because something is hard to work out!! And I learn a LOT along the way about what can and cannot be done.

 

In UV, we record a band of wavelength that is barely 50-60nm wide. Would it not be more logical to treat it all "together", as single "channel", by converting it to grey-scale, or, even better, using monochrome sensor for UV-photography?

 

(1) Oh, maybe. But unless we look at all aspects of it, we cannot possibly make that decision.

 

(2) For a long time now I have been looking at the raw colours of the UV photographs. It has become clear that we do indeed get two different raw "colours" in many UV photographs made with a broadband UV-pass filter and a lens capable of recording 300-400 nm. I just call the colours Magenta and Orange. Granted, those two raw "colours" are not very far apart, but the distinction is there. I see it over and over.

 

LINK: http://www.ultraviol...v-pass-filters/

This link shows how some filters are a bit more Orange and some a bit more Magenta. But with the BaaderU (just one example) I can get either depending on what I'm shooting using Sunlight or Sunlight+UV_Flash as illumination.

 

(3) As a function of the 3 human cone response curves, within a 60nm bandwidth we can and do get more than one colour. If you take yellow to be 580 nm, then 580+30=610 nm is orange and 580-30=550 nm is green. So perhaps we can record more than one "UV colour" within a 60nm bandwidth as a function of the camera's 3 response curves. If we had some detailed camera response curves between 300-400 nm to actually look at, then we would have some insight.

[Violet, blue and azure are within 60nm around 440nm.]

[Yellow, orange and red are within 60nm around 610nm.]

[Andrea B.]

I think you have yellow a little too low. 340nm is green.

 

Yes, perhaps.

But this rendering is dependent on white balance.

What does it look like in its original raw colours?

And how does the type of illumination affect it?

[Alex H]

Andrea, original raw colors in RAW-Digger do not account for the transmission of RGB dyes in UV part of the spectrum. If we do not know this transmission data, how can we interpret original colors in RAW-Digger data better than the properly white-balanced colors? White-balancing at least equalizes for the unequal transmission of RGB dyes across UV spectrum, and for internal boost of the signal in Red channel that all demosaicing algorithms perform.

[Andrea B.]

White balancing varies by camera and by app. So I just avoid it.

 

I do not claim that the interpretation of colour is correct from Raw Digger! Merely that by looking at the raw colour we have avoided the complication of imprecise results produced by the various white balance tools.

 

If someone would just sent me a spectroscopy rig and teach me how to use it, couldn't I solve all this in a flash?? Or am I being too optimistic?

 

The only way to know the original colour whether visible or impossible is by measuring the reflection off the subject. But all I have is a camera. Like others, I'm curious if that can be made to work. I've always said no, not accurately because of the many-to-one problem. But that is not to say that we cannot learn something from what we are seeing.

 

FWIW, you can look at the really raw file in Raw Digger without the consolidation of the two green channels.

[Alex H]

White balancing varies by camera and by app. So I just avoid it.

 

Are you sure it is white balancing and not built-in color profiling that causes variations?

[Andrea B.]

Andy, I realized I did not answer your question above in post #18.

Added an answer:

http://www.ultravioletphotography.com/content/index.php/topic/2102-flowers-with-two-false-colours/page__view__findpost__p__14892

[Andy Perrin]

Andy, I realized I did not answer your question above in post #18.

Added an answer:

http://www.ultraviol...dpost__p__14892

Ah hah. Ok, that makes sense, then.

[Andrea B.]

Are you sure it is white balancing and not built-in color profiling that causes variations?

 

It is, of course, probably a combination of both white balancing and colour profiling (and more). I've always thought that we should use both WB and colour profiles. One is really not good without the other. (ymmv....)

 

Although ---- in the newer cameras (as opposed to the olden days of the D70s and the D200s), both white balance and colour profiles have become very very good for visible light.

 

However --- when dealing with old lenses, converted cameras and ultraviolet false colours made under different UV-pass filters, a lot of that built-in accuracy falls by the wayside, methinks. There really are a lot of variables to control. I certainly get minor variations across my A7R, D600, K1 and GH1. And then we have some cameras which will not WB at all in the UV. And also some converter apps which cannot handle UV white balance. Those 2 factors add to the probem.

 

As always, standardization need only be "good enough" not perfect. At least for posting here on UVP. We are not a science lab, after all. Just interested hobbyists, I suppose. I'm not sure that I get paid anywhere near "enough" for my UV photos to call myself a UV professional over a hobbyist however passionately I may pursue this area of photography.

[Alex H]

Andrea, I am talking about camera profiling that is done by RAW converters automatically, "under the hood", and not by color profiling as you and Bjørn do using Color standards.

[OlDoinyo]

How do we know if a certain color in visible spectrum that we perceive (or camera records) is represented by light of a single wavelength, or by a combination of different wavelength? The same question that you always ask about false-colors in UV?

 

You are describing the phenomenon of metamerism: two distinct mixtures of wavelengths are called metamers if they produce the same tristimulus value at the sensor (or retina.) The answer to your question is that you don't, without resorting to apparatus with more channels (finer wavelength resolution.) I once demonstrated this to a group of schoolchildren with a prism spectroscope and two monitors (CRT and LCD) with color swatches tuned to appear identical to the eye. However, viewing the swatches through the spectroscope revealed very different spectral lines. First principles suggest that this must be possible with any multichannel sensor at any wavelength range.

[Alex H]

You are describing the phenomenon of metamerism: two distinct mixtures of wavelengths are called metamers if they produce the same tristimulus value at the sensor (or retina.) The answer to your question is that you don't, without resorting to apparatus with more channels (finer wavelength resolution.) I once demonstrated this to a group of schoolchildren with a prism spectroscope and two monitors (CRT and LCD) with color swatches tuned to appear identical to the eye. However, viewing the swatches through the spectroscope revealed very different spectral lines. First principles suggest that this must be possible with any multichannel sensor at any wavelength range.

 

That was a rhetorical question. For those who missed the point, I will try to explain one more time. One of the arguments that Andrea uses against using UV-color palette, which was shown by Klaus and confirmed by Steve using different approaches, is this phenomenon of metamerism. With the same success, this argument can be used against visible color palette, but I do not see anyone doing it. Why is that?

[Cadmium]

and for those of you who don't know about Klaus' UV color palette, here it is:

http://www.fotozones.com/live/index.php?/forums/topic/34297-vis-uv-zinnia-a-first-320-340nm-pattern-found/

[Andrea B.]

Alex: Andrea, I am talking about camera profiling that is done by RAW converters automatically, "under the hood"

 

Yes, raw converters do apply certain profiling and sharpening and other adjustments upon opening a raw file in the app. Raw Digger, however, does not apply any "under the hood" tweaks. The Dcraw app can also be used to extract the raw data from a raw file without any profiling or other.

 

 

Clark: You are describing the phenomenon of metamerism.

 

Ah, thank you! I need to make that word an active part of my vocabulary as I continue to learn and try to discuss all this.

 

 

Alex: One of the arguments that Andrea uses against using UV-color palette, which was shown by Klaus and confirmed by Steve using different approaches, is this phenomenon of metamerism. With the same success, this argument can be used against visible color palette, but I do not see anyone doing it. Why is that?

 

I argue for scientific caution. No one denies, least of all me, that wavelengths can be assigned an RGB value.

I merely point out, time and again, that METAMERISM is at play in all this. (Use the word 3 times and it is yours!) And also, that there is a large array of factors producing any colour in a digital camera. And another large array of factors producing colour in a flower petal.

 

 

 

Alex: With the same success, this argument can be used against visible color palette, but I do not see anyone doing it. Why is that?

 

If there is some app which purports to assign a visible wavelength to a visible RGB colour, then metamerism cannot be escaped there either. Whether or not that is brought to the attention of the users of such apps is something I cannot address unless I am familiar with the app.

 

BTW, no one here, least of all me, is "against" any colour palette.

So let's be careful how we word things, pretty please.

Urging scientific caution with this wavelength-to-RGB mapping, whether visible or uv, is simply common sense.

 

 

PIGMENTS

If someone could back up the wavelength-to-RGB mappings with a study of the underlying pigments in flowers/plants, then it would go a long way towards permitting some use of the digital camera in attempting to discern the reflected wavelength or metameric wavelengths. I have a book which presents a survey of plant pigments (given above as a reference), and the first thing I noted was that most flowers have a mix of pigments. OK, so, then we need to know how the pigments are mixed in the plant. Are two pigments literally mixed inside one chromoplast? Or are two pigments simply side-by-side in separate tiny packets? Either way, how does that affect the spectroscopy of a flower petal? And how does surface cell structure affect the reflection of these pigments? And so forth. There are literally gazillions of pigments across various chemical families, so it is daunting to attempt to figure out the wavelength-to-RGB value mapping even for a pure pigment.

 

---

Please keep the questions, comments, corrections coming!!!

---

 

 

3 LINKS to UV-to-False_Colour Maps

In addition to Klaus' palette (repeated here), we have a couple more to use as references.

 

Please notice right away that the UV wavelength-to-colour mapping in Shane Elen's link is completely different from the mappings in the other two links. And the two Panasonic palettes are similar but not exactly alike.

 

To record a mapping like this we must remember to specify each time camera, filter, lens, white balance used, method of recording, etc. We do not have that information for all these links. There is an extremely detailed 4-page discussion of gear, process and more in the link to Sascha's palette.

 

 

Shane Elen: 2006 Nikon D70

http://www.beyondvis...UVNikonD70.html

 

Klaus Schmitt: 2011 Panasonic GH1

http://www.openphotographyforums.com/forums/showpost.php?p=116037&postcount=2

But he removed this from his blog to re-write and apparently did not get back to it.

(Too busy I suspect! Happens to me too.)

http://photographyof...uv-palette.html

 

Sascha Hein: 2015 Panasonic G3

http://www.ultraviol...sonic-lumix-g3/

[Andrea B.]

I want to see how these wavelength maps look with no white balance, no sharpening, no profiles.

 

I want to see the Raw Data.

 

 

step one: order monochromator?

[Alex H]

In my personal opinion, as long as one does not know exact transmission properties of RGB dyes in UV spectrum, looking at RAW data with RAWDigger is misleading. This unknown factor (exact transmission properties of RGB dyes) can only be eliminated from the "equation" by white balancing on UV-neutral target.

 

 

I see no reason for myself to participate in this thread any longer. I wish the spring was here so I could actually take some UV-pictures with my new lens.

[Andrea B.]

Well, Alex, geez, no need to run off. It's important to get these insights and inputs. Else however will we learn?

 

I'll write to Daniel and Iliah at Raw Digger about it and see what they say.

 

And I still want to see the raw extraction of the Sparticle and the palette maps because I do not know what they look like in the raw. Do they show the progression of pink/red and magentas?

 

*****

What new lens did you get? Perhaps you can made do with some grocery store flowers??

[Andrea B.]

I have read only the abstracts. But these links serve to show that there is a lot of research on metamerism. Although the specific thing we are interested in is a bit narrower.

 

Information Limits on Identification of Natural Surfaces by Apparent Colour

Foster, Nascimento & Amano

Perception. 2005; 34(8): 1003–1008.

 

That one is about Visible colour.

 

 

Frequency of Metamerism in Natural Scenes

Nascimento & Foster

J Opt Soc Am A Opt Image Sci Vis. 2006 Oct; 23(10): 2359–2372.

 

Metamers arise because the number of degrees of freedom in the sensor system, three for the cones in the normal human eye or a typical camera, is smaller than the number of degrees of freedom needed to specify different spectra.

 

Again about visible colour.

[Alex H]

I am not running off, but if feels like we are going in circles here. I am 99.99% confident that what you will see in RAW pictures will be something very red or so. Red dye is made to transmit below 450nm so that RGB-based imaging process (sensors and monitors and eyes) can visualize those indigo and violet colors. So, red dyes do transmit a lot more UV than green and blue. That is why it needs to be compensated by proper white balancing, before any conclusions can be made. In my opinion.

 

New lens - UV-Nikkor. Grocery store flowers are boring in UV. In fact, flowers are boring in UV for me. When was the last time I posted flower pictures on this forum? (Flowers with pollinators is completely different story, but one needs to have both, and preferably alive).

[Andrea B.]

Of course we are going in circles. We lack data and experiments. :lol:

 

This unknown factor (exact transmission properties of RGB dyes) can only be eliminated from the "equation" by white balancing on UV-neutral target.

 

But I did that above!

I made my little colour diagram above by using both raw false colour and white balanced false-colour (to eliminate the Illuminant Factor) to show that there is a correlation. Possibly repeatable.

 

But none of that eliminates the fact that one wavelength in either UV, Vis or IR might produce the same raw false colour or white balanced false-colour as some other combination of wavelengths.

 

If only I had a 300-400 nm graph of my camera sensor response.....................

 

U R rite tho.

Time to go do something else for today. B) B) B)

 

---

Congratulations on your UV-Nikkor!! I hope you will enjoy it. Sharp as a tack. Easy to use. Get a PN-11 if you don't already have that.

[JCDowdy]

We lack data and experiments.....

 

I have been enjoying this mental floss for the last few days, so far resisting the temptation to kibitz! B)

 

But... we do have data. It has been shown and independently confirmed by users of the "Sparticle" bandpass filter arrays that one can record a spectral UV-yellow with a UV-white balanced Badder-U or similar kit.

 

Correctly stated, several different ways, is the fact that we do not know if a metameric UV-yellow can also occur.

 

That should not be a difficult experiment to conduct. Rather than photograph the array, why not aim it at an amply sized slab of PTFE or Spectralon and photograph the reflection?

 

The 350/360nm filters should throw a UV-yellow spot on the Spectralon target. However, if the target is positioned far enough from the filters, some of the beams should begin to overlap. Overlapping beams of spectral UV-colors might produce metameric UV colors. The filters in the "Sparticle" array might have to be rearranged to test all combinations but the results could be very interesting.

[DaveO]

Here's an interesting paper on flower coloration:

 

 

How to colour a flower: on the optical principles of flower coloration

http://rspb.royalsoc...3/1830/20160429

 

The anatomy of flower colour

https://phys.org/news/2016-05-anatomy.html

[Cadmium]

Not exactly related, but here is an old test showing an example of PTFE used as the illumination background for a small Sparticle.

It is not as bright, but uses the same PTFE for illumination that the white balance is made from, all of which derives its light from indirect sunlight.

 

Compare to this (upside down) which uses the sky as background illumination.

[Andrea B.]

John, we lack data about sensors' RGB response curves between 300-400nm and about whether UV induced fluorescence mediates UV response. And one or two measurements without a full description of methodology does not a certainty make. Which is not to say that it won't work out. I'm sure it will.

 

But -- show me *everything* including the raw data. Klaus for example never described his equipment or methodology. Sascha and Shane did (to different degrees). But the results in the 3 color palettes linked above are not the same! I suspect that is due to the use of different white balance steps, but I do not know. Shane's palette looks like raw false colors -- the same orange, pink, red tones I've described here and in other threads (linked above). Sascha's palette looks like white balanced false colors. Sascha's and Klaus' palettes are similar but not the same.

 

We also lack data about how the palettes hold up under different illuminants. If the existing palettes were "made in the dark" using a monochromator to transmit UV wavelengths, then does the palette shift when shooting scenes in sunlight or under xenon flash? If so by how much? In which direction?

 

In all my years of shooting UV with UV-Nikkor or CO60, I've never gotten a turquoise colour (Panasonic GH1 palette). And very very rarely have I gotten a dark green. I **cannot** get a false green (Sparticle palette, Panasonic palette) with my Edmund narrowband 340 nm filter, just the same old blues and yellows. I never get the Panasonic purple between 380-400nm. I get instead a much bluer color with only hints of violet. So does the Nikon D600 somehow produce a different palette? We have been assuming that all Bayer filters are pretty much the same in how they treat UV, but is that valid? We have no data to help decide that.

 

Too many questions, not *enough* data, an unsolvable mystery (Bayer dyes) and not enough money for me to indulge in the purchase of a spectro kit.

 

Oh well.