Flowers with two false colours

[Andrea B.]

Using a camera with a Bayer filter, we sometimes see two colours, blue and yellow, in white-balanced, false-colour UV floral signatures.

 

Some examples:

• Phlomis fruticosa [Jerusalem Sage]
  
• Iris pseudacorus [Yellow Iris] and Iris pseudacorus [Yellow Flag]
  
• http://www.ultraviol...-danfords-iris/)
  
• Brassica rapa ssp. campestris [Wild Turnip]
  
• Brassica rapa [Field Mustard]: Another Example
  

Here's a repeat of the Jerusalem Sage flower in Visible and in UV light. The UV photo was made in Sunlight with a UV-flash boost. The BaaderU UV-pass filter was used.

 

 

When what to our wondering eyes should appear but two false-colours......

 

 

Raw Composite of an Ultraviolet Signature

Here is a raw composite of the preceding UV photo extracted using the Raw Digger app. No white balance has been applied to this photo, which has been demosaiced, gamma-ed, stretched and saved as a Tiff (and then as an sRGB Jpg for posting here). I wanted to look at the raw colours and attempt to map them to the false colours in the finished UV photo above.

 

 

From sine waves to 8-bit RGB

The reflected UV light off the flower passes through a UV-pass filter, through a lens, through a Bayer filter and is collected into sensor wells. The sensor wells are read and the light is converted to a digital sample. The digital sample is converted to an 8-bit RGB colour represented in a colour space in the camera, a colour space in the converter/editor and then in a colour space on your monitor. So if I discuss the raw colours and the eventual corresponding false colours in the two images above, please know that I'm aware of all the twists and turns along the way (...well, I hope!...). And it is indeed a long, long way from the original reflected light to the final white-balanced, false-colour UV-signature photograph.

 

Raw colour to false-colour diagram

Below is a little colour diagram I made. But I have to explain it in the next two sections.

[Note: I forgot to put degree symbols on the colour diagram.]

 

10° --> 235°

The raw colours which eventually become false-blue are on the colour wheel around 8°, 9°, 10°. They vary in their brightness. I sampled one of the 10° colours, (101,62,54), and placed that colour in the 2nd box from the top. A fully saturated 10° sample colour is placed in the 1st box from the top.

 

The 8°,9°,10° raw colours all mapped to false-blues around 235° in the colour wheel. A sample 235° false-colour from the finished photo, (99,105,170) was placed in the background behind the two 10° colours.

 

 

16° --> 52°

The raw colours which eventually became false-yellow are on the colour wheel around 14°, 15°, 16°. They vary in their brightness. I sampled one of the 16° colours, (186,111,83), and placed that colour in the bottom box. A fully saturated 16° sample colour is placed in the 3rd box from the top (2nd from the bottom).

 

The 14°, 15°, 16° raw colours all mapped to false-yellows around 52° on the colour wheel. A sample 52° false-colour from the finished photo, (221,216,184) was placed in the background behind the two 16° colours.

 

 

Conclusion:

I don't mean to let you down with a thud, but I don't have a conclusion.

What I find interesting is that two sets of raw colours which are about 6° apart on the colour wheel are mapped to two sets of false-colours which are around 177° apart on the colour wheel, that is, almost complementary. Applying white balance to the raw colours really does stretch out the relationship between the corresponding false-colours. [i suppose we should call the white balance step a "false-white balance" step.]

 

Question: How in the world does any of this relate to the original wavelengths? If (as some think) false-yellow indicates a shorter reflected UV wavelength than does false-blue, then what wavelengths do all the false-grey areas in the photo represent? Can two raw colours which are only 6° apart on the colour wheel really represent shorter and longer UV wavelengths as passed through a BaaderU? I suppose so, but something about all this seems a little shaky.

 

Added: 02 February 2017

There is obviously a correlation. I didn't mean to imply there is not.

 

[Andy Perrin]

It's surprising (to me) that multi-UV-false-color flowers don't happen more often. In visible light, flowers with, say, yellow and blue on the same petal even occur pretty frequently. Iris versicolor is pretty extreme in that regard, but there are plenty of others. Interesting that it's again the irises that you chose for your example...

[Alex H]

If (as some think) false-yellow indicates a shorter reflected UV wavelength than does false-blue ...

 

Are there any compelling arguments against this theory?

 

... then what wavelengths do all the false-grey areas in the photo represent?

 

What about "real" "visible" grey in "visible" "normal" pictures? What wavelength does it represent?

[Andrea B.]

Hi Alex. :)

 

What I'd like to see is spectrographic proof. Just once. If I had (US)$10000, then I might try to measure some of this stuff myself. But I don't think I want to spend that kind of money in order to wander around measuring flowers and leaves. :D

 

In visible light the RGB white or grey areas represent reflection (or absorption) of equal amounts of R, G and B.

 

In UV-only, I suppose we can say that false-grey areas are partially absorbing (or partially reflecting areas)? But false-blue and false-yellow areas are also partially absorbing/reflecting. So false-grey must be partially absorbing/reflecting both of the wavelengths* which cause false-blue and false-yellow? If indeed false-blue and false-yellow arise from a single wavelength*. There could just be reflected additive colour behind the false-blue and false-yellow.

 

*ADDED: That was stupid. Any reflected colour off a flower consists of a range of (additive) wavelengths. No pigments I've ever seen can be correlated with a single wavelength. I should have been writing wavelength range. False-yellow and false-blue arise from some reflected wavelength bands.

 

OR -- it might be that due to the Bayer filters, camera white point and white balance algorithms, the grey areas actually represent a wavelength-based reflected colour of some type.

 

I need to go look at more raw files in Raw Digger to see if there is ever any grey in them. I don't think so because all the raw histograms I've extracted so far seem to be predominantly red on the small sampled areas with smaller amounts of blue & green. But then I have not exactly conducted a through study of this. Because it all leads back to simply wanting spectrographic results.

 

I lack the proper vocabulary here.

A laser transmitted colour -- single wavelength -- is one thing -- light passing through something.

A reflected colour is another thing. It may be additive and not representative easily correlated with a single transmitted (wavelength) colour -- would need a wavelength range to describe it*?

 

*ADDED: I added the phrase beginning with "would".

 

So attempting to trace reflected, additive colour (false, no less) back to some correspondence with a transmitted wavelength range* is fraught with outright PERIL, imho.

:D :D :D

 

*ADDED: the word range.

 

EDIT: I tried to cross out some stupidity. I lack the vocabulary for this, but I'm attempiting to develop it.

[Alex H]

So how do we know if visible colors are additive or not?

[Andrea B.]

? not sure what you are asking ?

 

Relected light is additive.

[Hornblende]

Do we know the transmission curve in UV of the red, blue and green filter over the pixels?

[Alex H]

? not sure what you are asking ?

 

Relected light is additive.

 

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?

[Alex H]

Do we know the transmission curve in UV of the red, blue and green filter over the pixels?

 

Such data exist but was not published.

[Andrea B.]

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?

 

We don't know. As far as I know. So we don't know about that in UV light either. Which is why I always worry about trying to step backwards from false-colours to a wavelength -- or a wavelength band - or a finite union of small wavelength bands. It may be that false-yellow derives from shorter UV wavelengths. But what if false-yellow -- in a similar manner to visible yellow -- derives from two separated wavelength bands one of which is shorter wavelengths and one of which is longer wavelengths -- similar to how visible yellow can be the addition of green and red, green wavelengths being shorter than red wavelengths.

 

Ok, enuf. I'll let others try to figure this out. I need to perform some website duties. :D

[Andy Perrin]

Andrea, try this argument: Suppose we know that the post-WB UV yellow produced by a laser is at 340nm. Then any mixture that produces an equivalent yellow must have some wavelengths on each side of 340, even if it contains no 340 at all, like, say 335 and 345. So that means 335 (or some number <340) is present!

[DaveO]

Here's my contribution to muddying the waters

 

From my darkroom days, although I was never a colour printer, I seem to remember that blue and yellow are complimentary colours so that the more blue light you have the less yellow when thinking of the yellow-blue axis. That's how the the contrast of "multigrade" B&W papers works, one of the layers, say the high contrast one, is sensitive to blue light so if you use more yellow light more or less of that layer is exposed.

 

The problem with trying to find the UV absorption curves of the Bayer dyes (don't worry I have tried but I only have the free databases to go on) is that they are almost certainly proprietary dyes so their properties will never be revealed anywhere I can see. We can assume they would be azo dyes for their light stability but that doesn't get you very far. The other problem is that the Bayer array was only ever designed for the visible region of the spectrum.

[Andy Perrin]

The discussion of dominant wavelength is pretty interesting...

[Andrea B.]

^{Andy: Suppose we know that the post-WB UV yellow produced by a laser is at 340nm.}

^{Then any mixture that produces an equivalent yellow must have some wavelengths on each side of 340,}

^{even if it contains no 340 at all, like, say 335 and 345. So that means 335 (or some number}

 

^{The mixture which produces a reflected colour equivalent to a spectral colour may not necessarily consist of wavelengths from each side of the spectral wavelength. For example, take spectral violet at 400nm. Violet can be also created also with a mixture of blue and red light both of which are to the right of 400nm, namely (131,0,141). So we have in this odd case a shorter wavelength spectral colour being also created with a mixture of longer wavelength colours. Weird, yes? (Did I do this right??)}

 

^{Now, I grant you that this example is entirely dependent upon the way our eyes (and similarly our cameras) process visible light because the response curve of our red-detecting cones has a very long left-hand tail which extends into the response curve of our blue-detecting cones thus creating a response of B+R whenever spectral violet enters our eye.}

 

^{Can such a thing happen with UV light? We could only answer that question if we have information about the camera response curves between 300-400nm.}

 

^{Ref for human cone response curves:}

^{Livingstone, Margaret (2014) Vision and Art: the Biology of Seeing. Abrams, New York, NY.}

 

---

 

^{Dave, I've searched more than once for information about the Bayer filter dyes but have never found much.}

[Alex H]

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?

 

We don't know. As far as I know. So we don't know about that in UV light either. Which is why I always worry about trying to step backwards from false-colours to a wavelength -- or a wavelength band - or a finite union of small wavelength bands. It may be that false-yellow derives from shorter UV wavelengths. But what if false-yellow -- in a similar manner to visible yellow -- derives from two separated wavelength bands one of which is shorter wavelengths and one of which is longer wavelengths -- similar to how visible yellow can be the addition of green and red, green wavelengths being shorter than red wavelengths.

 

So, shall we exercise the same precaution for colors in visible spectrum? Shall we, when talking about colors that camera record, always assume and state that "yellow" flowers may not necessarily reflect narrow "yellow" wavelength, but a combination of other wavelength, and so on for other colors?

[Cadmium]

By using narrow bandpass filter, you can define a color for a wavelength as it appears to your Bayer filters depending on white balance.

For example, if you use a 435BP10, your Bayer would see blue light of 345nm +/- 5nm.

The filter on the lens defines the overall band range for the Bayer, the bandpass filter defines a narrow segment of that overall band.

I know you don't like it when I pull out my Sparticle, but it shows what the Bayer filters see per narrow bands of the spectrum, depending on white balance, but that is up to you how you white balance it.

Sorry for the pics, again, but I see no other way to illustrate my point.

 

UV-A

 

Visual

 

IR

[Andy Perrin]

I have no idea what would be wrong with using the Sparticle? That is indeed the right way to find out what false UV colors are related to each range of wavelengths (for the given WB). I think we all understand that you can't reverse the mapping from some reflected mixture back to a single wavelength. But the other direction (wavelength->colors) is clear. Cadmium, can you tell me what the actual filters are in that array so I know what wavelengths go with each?

 

One thing that sticks out at me here is that the ordering of the false colors by increasing wavelength does not follow the classic ROYGBIV (or VIBGYOR in this case) order. So if we wanted to imagine what our UV photos look like if we simply moved all the wavelengths down into visible (i.e. map 300-400nm to 400-700nm), we should be rearranging the ordering of some of these colors.

[Andrea B.]

---

 

Alex: So, shall we exercise the same precaution for colors in visible spectrum? Shall we, when talking about colors that camera record, always assume and state that "yellow" flowers may not necessarily reflect narrow "yellow" wavelength, but a combination of other wavelength, and so on for other colors?

 

That's always what I have understood to be true. Indeed, while reading about floral pigments last night I ran across the statement by the author that she has seen no spectral violet pigments or objects which are not due to diffraction or interference.^{(1, pg 44)} So what we perceive as a violet flower is almost always a combination of reflected red and blue. Individual floral pigments do have a wavelength peak in their reflectance profile, but most flower colours are mixtures of pigments so that the reflection is additive.^{(2, pg 76)}

 

I think we have to keep in mind that we are dealing with the quirks of our human visual response curves when attempting to make any statements about reflected visible color. That is to say, when we perceive a single wavelength yellow, the wavelength is actually triggering an +R+G response in our eyes due to overlap of the red and green human response curves. [Note: Strictly speaking the human response curves are not "blue", "green" and "red", but we refer to them colloquially like that.^{(1, pg 39)} ]

 

As to what we actually state when talking about the camera record, well I will leave that up to everyone to decide how they wish to handle that.

 

Reference:

(1) Livingstone, Margaret (2014) Vision and Art: the Biology of Seeing. Abrams, New York.

(2) Lee, David (2007) Nature's Palette: the Science of Plant Color. The University of Chicago Press, Chicago & London.

 

---

 

Steve: I know you don't like it when I pull out my Sparticle

 

Steve, not true!! We love the Sparticle!!

 

OK, so, we know that, for example, the 4th filter from the left triggers a yellow response (+R+G-B ) from the camera due to the peak wavelength that particular filter passes. But there may also be additive combinations of UV wavelengths which trigger that same response. That is what we do not know. Are there such combinations in the UV? If we had a very detailed spectral response for the 300-400 nm range in one of our cameras, we could then possibly determine the answer to this by examining the overlaps in the R, G and B responses in the UV range. Once we knew how the camera response worked for potential additive combinations in UV, we could perhaps set up illumination and filtration experiments to trigger it.

 

---

 

Andy: One thing that sticks out at me here is that the ordering of the false colors by increasing wavelength does not follow the classic ROYGBIV (or VIBGYOR in this case) order.

 

That is true of the white-balanced false colours.

You need to look at the raw false colours after demosaic before white balance.

The photograph of the sparticle can be converted in Raw Digger for this.

 

EDIT 01 February 2017. That is an opinion! I think it is better to look at the colours before white balance is applied BECAUSE each camera and each app applies white balance just slightly differently.

 

---

 

oh mercy me. I so desperately need a spectrometer, monochrometer and so forth. Maybe I'll take out a loan. And then take a class about how to operate it all. sigh.

 

so if you saw a grey-haired lady running around the park pointing a wand at all the flowers, what would you think?

:D :D :D :D :D wheeeee.....unicorns!!

[Andrea B.]

ADDED: Always stand ready to be corrected when I get it wrong! Otherwise I'll never learn.

[Cadmium]

Andrea, Sorry, I was going to edit that out, no big deal, I have had people question the Sparticle idea in the past, no one person, so just a joke.

Enrico uses the idea also, or at least he use to.

 

Andy, Here is a labeled version, but I have changed a few of the BP fillers over time. For example in this one the 460DF20 has a known IR leak, thus it looks a little like the GG420 sometimes.

[Andy Perrin]

Andrea: why look before the WB if our WB really is standardized (i.e. two cameras with the same lens and filter and sensor glass should make the same colors after WB)? It should work either way, even though the actual colors will be different ones. We only need consistency here. The colors in our photos are all post-standard-WB, so I think that's what we should stick with.

[Andrea B.]

I suppose I simply keep it all up front so that we keep in mind that we are dealing with a LOT of "translations" when we go from UV wavelengths to visible false colours. There is a lot of hand-waving (assumptions, approximations) in this process and we need to keep in mind where it is occuring. And look at how it happens.

 

 

EDIT 01 February 2017

To answer Andy's question, I avoid looking at the white balanced colours in this kind of analysis because the white balance tools in cameras and in apps each produce very slightly different results. Go look through the botanical posts all of which have been "standardized" across the various platforms. The blues are not all the same. The yellows are not all the same. Some WB apps create a bit of dark green; others do not. The results are standardized enough, but the standardization is by no means always accurate.

[Andrea B.]

Looking at the raw and white-balanced colour samples from above.

 

Blue false colour

 

UV source, if single ~= 380 nm (? guessing)

 

Raw sample (101,62,54) under illumination of Sunlight + SB-14 UV-flash

HSB = (10°,47%,40%)

(10°,100%,100%) = (255, 41-44, 0) ~= 638 nm

 

WB sample (99,105,170)

HSB = (235°,42%,67%)

(235°,100%,100%) = (0,20-23,255) ~= 442-443 nm

 

Yellow false colour

 

UV source, if single ~=340 nm (? guessing)

 

Raw sample (186,111,83) under illumination of Sunlight + SB-14 UV-flash

HSB = (16°,55%,73%)

(16°,100%,100%) = (255,66-70,0) ~= 632-633 nm

 

WB sample (221,216,184)

HSB = (52°,17%,87%)

(52°,100%,100%) = (255,219-223,0) ~= 590-591 nm

 

So - under a lot of assumptions and approximations -

• 340nm UV, recorded as less intense 632nm, white-balanced to less intense yellow 590nm, and
  
• 380nm UV, recorded as less intense 638nm, white-balanced to less intense blue 442 nm.
  

OK, so now I'm temporarily brain dead and must stop for a while. B)

I don't think I can ignore the additivity like I have done.

I'm not sure I can assign the wavelengths like I have done.

[Alex H]

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?

[Cadmium]

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

 

If you are using a 300(320)nm - 400nm lens filter, then

UV Blue = 400nm

UV Lavender = 385nm

UV Yellow = 350/360nm

UV Green = 340nm

UV Turquoise to Cerulean = 320nm to 300nm

 

Now, if you are using a lower lens filter, that restrains the upper UV-A range, you will get some different color balance, depending on your lens, light...