Why Is It So .....

[colinbm]

Why Is It So .....
That when we take a White Balance on PTFE/Teflon, that is nearly 100% reflective, it is expected that the PTFE/Teflon will be white in the scene ?
If I shine a red light on the PTFE/Teflon, then it will be red.
If I shine a green light on the PTFE/Teflon, then it will be green.
If I shine a blue light on the PTFE/Teflon, then it will be blue.
Now of course, If I shine a white light on the PTFE/Teflon, then it will be white.
Easy-peasy...

But, If I shine a UV light on the PTFE/Teflon, then it should reflect the UV. 
Which it should be the CFA equivalent of how the dyes respond to the UV ? 

Yes / No ?

[lukaszgryglicki]

Hmmm I think yes.

[Stefano]

I don't think I understand your question.

 

PTFE is white in UV like a white surface is white in visible light, meaning it will reflect all UV wavelengths well. I don't understand the CFA part. The false color you get is the sum of all the UV wavelengths the sensor sees with their false color and weights (an integral essentially), which is analogous to visible white light, although the CFA doesn't separate UV colors as well as visible colors, because it wasn't designed for that.

[lukaszgryglicki]

But it shoudl give an answer on how "white" UV records in the camera - like ho CFA records UV when all UV wavelengths are present/reflected.

So when you "white balance" to it then somethign white/grey in UV is white on the photo...

 

 

[Kai]

I'm also not sure if I understand your question correctly.
So here are a few statements that I hope will help answer your question:
 

Next to the PTFE
It reflects only about 80% light, depending on the surface structure. But that's no different with an aluminum mirror behind glass (unlike many assume). White snow: Also about 80% or less. The only important thing is that PTFE reflects equally well over a very wide range of wavelengths.
https://en.wikipedia.org/wiki/Albedo
 

Now to the dyes (CFA)
Of course, they also have an absorption spectrum in the UV and in the IR. In the NIR beyond 850 nm, the absorption curves no longer differ. This means that the sensor cannot display any false colors for wavelengths beyond (approx.) 850 nm (i.e. no longer distinguish between the wavelengths).
In the UV it's a little different.
Attached is a picture to illustrate a specific example.
I recorded the spectrum of sunlight through a Schott DUG11X and an enlarger lens. The sensor still has the Bayer mask, but no other filters.

 

The upper spectrum has no special WB but the camera setting "Daylight". The lower one is balanced on PTFE.

The three curves on the left now show the reaction of the red (top), green (middle) and blue (bottom) pixels along the image (left edge 325 nm, right edge 400 nm).
You can see that the red dye transmits the most UV - regardless of the wavelength. The green dye only lets light through below 380 nm, the blue dye blocks below 365 nm.
Depending on the monochromatic wavelength (which hits the sensor), a mixed color results.

 

With a WB on PTFE (the right row of spectra), I set the color sum of all registered light to zero. Then the red channel is weakened, the green and the blue are strengthened. The sum of the three curves on the right is therefore "white". Since the red and green channels now practically no longer differ, they together result in yellow and in our UV false color images we have the color spectrum between yellow (short-wave) and blue (long-wave). In the middle, RGB are equally strong and it's "white" right there.
Since the red and green curves are not exactly the same, the yellow appears greenish or even green in the very short wavelength range.
There are, of course, subtle differences between the individual sensors/dyes used.

I hope this helps you, Colin.

 

[lukaszgryglicki]

Great summary, thanks for this!

I wonder how it looks like in 200-320nm range (if coverglassis just quartz passing from 180) - I know that CFA is probably pitch-black there but if we"force" enough light there, will there be any more differences between R, G, B dyes, somore false colors could exist in UV-B and UV-C?

I guess nobody ever tried to measre this, because that deep in UV we usally use mono.

 

[colinbm]

Thanks for the comments.
I guess what I am saying is, that the PTFE/Teflon is acting like a mirror & reflecting what we shine on it, so why do we WB off it in a UV only image ?
Should we be leaving it as the UV yellow & green that the CFA gives it ?

Kai has mentioned on another question I had for later & that is, has anyone seen graphs of the responses of the CFA lower then 300nm ?

[ulf]

24 minutes ago, colinbm said:

Thanks for the comments.
I guess what I am saying is, that the PTFE/Teflon is acting like a mirror & reflecting what we shine on it, so why do we WB off it in a UV only image ?
Should we be leaving it as the UV yellow & green that the CFA gives it ?

Kai has mentioned on another question I had for later & that is, has anyone seen graphs of the responses of the CFA lower then 300nm ?

The CFA (Bayer matrix) normally creates a very big unbalance in the three channels, at least with a wider band UV-source like an Uv-modified flash or with sunlight.

For typical filers and filter combinations with "Baader U style" the raw images will be very RED.

 

With some other stacks like a BUG-stack things are even more extreme, but then WB against PTFE will not give the most pleasing colour palette.

There I normally tweak to taste.

 

Just as there is a need for different WB in VIS images taken in sunshine or candle light there is a need for WB with wide band light sources.

 

This fact will not apply when using narrow band light sources.

[Andy Perrin]

1) There is no such thing in nature as “white” or “grey” — humans define it. 

2) Almost anything can be made white/grey in a photo by white balancing off of it. 

3) Under sunlight at noon, the objects that humans define to be white reflect a spectrum that more or less matches the spectrum of the sunshine that hits it. This is the D65 spectrum (more or less — they actually use a blackbody of a certain temperature as a stand-in for noontime sun).

 

4) From the above description, you can see that the definition of “white” requires visible light, since the D65 spectrum is visible, and that is our *definition* of what we mean by white.

5) In UV we are free to make up our own definition of white, and a convenient one is the reflection of your chosen light source off PTFE, which reflects the color of your source, and therefore removes the color cast we otherwise see in UV photos. 

[Kai]

4) It's correct. It is about a term that compares stimulus (light) and receptor (eye, sensor). We call a specific combination of receptor stimulation "white".
But: By analogy, in acoustics one also speaks of "white noise" (and "pink noise" when all frequencies/wavelengths are present with the same intensity, also analogous to the situation of light/human eye).

 

5) It's correct. We are free. That's why I like to flip the color spectrum in my UV "color" pictures. Then short-wave UV light looks blue/violet and long-wave UV light looks yellow. This corresponds in analogy to the VIS range. A purely subjective decision. (Maybe I should swap yellow for red? That would be even more consistent in this sense.)
There is a similar discussion at the "Bee Vision". Do you leave "blue" and "green" as they are and set "UV" equal to "red" or do you get the order according to the wavelength? Then UV=blue, blue=green and green=red. The colors then differ, but the color balance is not affected by this :)

[lukaszgryglicki]

Actually I would like something very simple:

Divide frequency by 2 and display with R, G, B, so.

200-240nm will become blue (just like 400-480)

240-300nm will become green (just like 480-600)

300-400 will become red (just like 600-800)

 

That would be very interesting, or just UV-C -> blue, UV-B -> green, UV-A -> red.

 

Same can be funny or multiply by 2 (IR)

750-950nm will become blue (just like 375-475)

950-1200nm will become green (like 4875-600)

1200-1600 will become red (just like 600-800)

 

Or just NIR -> blue, SWIR -> green, LWIR ->blue.

 

Would love to see such images.

[Andy Perrin]

Lukas we have done these kind of images many times here. Please see Bernard’s work and Stefano’s. 
 

LWIR is a bit complicated because you see partly emitted light and partly reflected, whereas the NIR and SWIR are only reflected (unless temperature is extremely high). That would make a tri-color hard to interpret. 

[lukaszgryglicki]

UV (C, B, A)? wow - never seen one there (combined in one image) - I'll search...

I think I've seen different IR bands combined into RGB (but all below 1.2 um) but I don't remember seeing (LWIR,SWIR,NIR)->(R,G,B) - order RGB doesn't matter - what matters is combination of LW, SW and NIR in one image.

[Andy Perrin]

Yes, I think david did A,B,C. Also Jonathan.

 

i just explained that LWIR has problems for these so I will not be doing it! Please read. 
 

I did NIR/SWIR myself (780, 1064, 1450 if I recall).

—-

This thread was about white balancing and I don’t know why you are changing the topic even. 

[lukaszgryglicki]

Yep, LWIR will emit others are reflected - understood.

[Andrea B.]

If I shine a UV light on the PTFE/Teflon, then it should reflect the UV. 
Which it should be the CFA equivalent of how the dyes respond to the UV ? 

Yes / No ?

 

No.

In the visible world, we do not define "white" as having anything to do with the CFA of our digital cameras. We define "white" using physics, i.e., using the properties of electromagnetic wavelengths. So, similarly, in the ultraviolet world, we should not use CFAs to extend the definition of  "white".

 

<Added Later>

After reading Andy's response below, I think it would be better if I wrote the first paragraph as follows:

One way we define "white" is by using physics, i.e., using the properties of electromagnetic wavelengths. The other way we define "white" involves our eye/brain perception. Neither of these definitions involves our camera CFAs.

<Added Later/>

 

Here's what I've written before (more or less):

Given that "white" is defined in the visible world as that property of an object which reflects all visible wavelengths,

then the reasonable extension of that concept would be to say 

"white" is that property of an object which reflects all visible and all ultraviolet and all infrared wavelengths.

This is exactly what PTFE or Spectralon or XYZ does. 

(Well, at least in the range from about 300 - 1000 nm, or so.)

 

And please remember that the CFAs are not the same in every camera. Prime example currently being that Panasonic S1R which I have been wrestling with. The UV photos from the S1R don't look quite the same as those from the converted cams we have all used so far. So using a CFA response to define the "white" property would not provide a stable definition.

 

You'll have to decide for yourself whether it is indeed reasonable to attempt an extension of visible white to UV or IR wavebands. The only reason we did that here on UVP was to provide for a uniform presentation across gear platforms in the Botanical Section by applying a white balance based on a PTFE/Spectralon/Other standard. And it does help tame the riotous UV false colours which can occur otherwise.

 

[Andy pretty much had this above. But I would disagree that nature doesn't have "white" or "grey". It certainly does. I think Andy probably meant that there is no white wavelength or grey wavelength. Or black or magenta wavelengths, either. So we have to define the white/grey/black/magenta concepts in terms of other wavelengths and wavelength intensities, etc.]

 

 

 

Hey, feel free to disagree, debate, protest or maybe even agree. I don't get upset if I'm all wrong about stuff. Or partly wrong. ....or whatever. 

[Andy Perrin]

Quote

[Andy pretty much had this above. But I would disagree that nature doesn't have "white" or "grey". It certainly does. I think Andy probably meant that there is no white wavelength or grey wavelength. Or black or magenta wavelengths, either. So we have to define the white/grey/black/magenta concepts in terms of other wavelengths and wavelength intensities, etc.]

No, I really meant that nature doesn't have white, because "white" is a thing that ultimately happens in our heads! Our vision tends to adapt to the light we are in (like our own internal white balance) and also adapt to the colors in the scene, etc. We mentally make these adjustments for the spectrum of the light we happen to actually be in at the time. If it's twilight, a color that would otherwise appear very blue is "white" to us. So it is what I said, for monitors and things, we DEFINE white to be the spectrum of a blackbody that's pretty close to noontime sunshine. You can buy a lightbulb that approximates that spectrum, and it will look reasonable in the daytime, but then when night comes, it will seem too darn blue-ish and you'll wish you had something with a color temp that isn't that standard 6500K of the noontime sun - people like warmer color temps at night. Colors are in our HEAD, which is a point you have made yourself, Andrea, many times.

 

Anyone remember The Dress?

https://www.thesun.co.uk/living/3496637/optical-illusion-dresses-colour-change/

 

Objects that have a constant reflectivity in visible can be said to be white or grey, and it's reasonable to extend that to other parts of the spectrum (Andrea, we agree on this for sure).

[Andrea B.]

Actually we agree on all points. I do get what you are saying. 

 

To clarify a bit -- that is, to split a few more hairs.... 

 

I was trying to define "white" as a property of an object having certain measureable, electromagnetic reflectivity characteristics. 

 

But "white" can also be defined as a color of an object in the sense of stimulating a particular eye/brain perception -- where that perception certainly might vary, as you note, depending on what the object is next to, what illumination the object is under, and/or how our individual eye/brain connections work when the reflected wavelengths hit them. 

 

BTW, I've seen arguments that white is not a color, per se, but that's splitting the hairs so finely that the bunny is going bald. 

 

[Andrea B.]

Kai, I liked your post about the Bayer dyes and the explanation of white balance. Very helpful to keep in mind always how that works.

 

Ulf, that is a very good point about narrowband light sources -- applying white balance usually doesn't work well.

[Andy Perrin]

46 minutes ago, Andrea B. said:

Actually we agree on all points. I do get what you are saying. 

 

To clarify a bit -- that is, to split a few more hairs.... 

[snip]

BTW, I've seen arguments that white is not a color, per se, but that's splitting the hairs so finely that the bunny is going bald. 

Poor bun! Yeah, I mean, white isn't a spectral color, but lots of colors (pink, brown, etc.) are not spectral colors, and that wasn't what I was getting at.

--

 

Now that we got that figured out, some further thoughts:

 

The reason white light is defined to have a specific color spectrum is to allow comparisons between colors in the same object under photographic conditions, or on displays for example, or between fabric swatches or paint chips, etc. It prevents disagreements like The Dress. One thing we do not have is an equivalent for the D65 illuminant in UV. That is, we are all using many different types of UV light source with varying illumination, different amounts of each wavelength, etc. At least some part of the issue with have comparing false colors (and this would not make them any less false!) is that we haven't got a standard illuminant — the closest thing is the sun, but of course that varies with time of day just as with visible. If we ever wanted to get serious about examining floral false colors scientifically or something, I think that would have to be a necessary first step. Obviously it doesn't matter in artistic photography.

[colinbm]

Thanks for the discussion & inputs.
I am learning.

[Andrea B.]

Andy, yes, there are for sure a lot of problems to be solved in trying to extend color analysis to the false colors of UV. 

The big problem to me has always been the Bayer dyes. They are meant to generate human visible colors. But as Kai has discussed above, the Bayer dyes don't "nicely" pass the UV in such a way that the low third, middle third and high third of the waveband 300-400 nm each pass through a separate dye. We can only produce such an equivalent by using narrowband filters and narrowband UV LEDs to make TriColor photos. But we don't have narrowband filters and LEDs that neatly divide the UV waveband so I always feel like something has been left out somehow.

And so on......

 

The D65 illuminant definition could probably be extended to UV because it has a basis in natural sunlight? But actually providing any kind of standard UV illumination is a whole nother ball of wax. Every kind of UV illumination that we have omits some parts of the 300-400 nm region. 

 

I do not know enough about illuminants and color spaces and all that to contribute anything else to this interesting discussion. So I'd better stop here and let other members weigh in. 

Lots of fun to think about all this. Wish I did know more.

 

 

 

 

 

 

[Andrea B.]

P.S.  Another comment about Bayer dyes...

Remember a few years ago OlDoinyo discussed the possibility that fluorescence of the Bayer dyes mediated the false color response? Did that ever get settled? Do the Bayer dyes fluoresce? 

[colinbm]

I have not seen the Bayer CFA fluoresce, but that is through the cover glass that should allow the UVA light in & the fluorescent light back out ?
I would like to see a graph of the Bayer CFA that reaches below 300nm like it would have with a fused silica window.
Here is one that I have found ....
https://www.flir.com.au/support-center/iis/machine-vision/application-note/understanding-color-interpolation/

[colinbm]

This one has a better response at 300nm 
https://www.researchgate.net/figure/Color-pixel-quantum-efficiency-of-OV9715-RGB-CMOS-sensor_fig11_277504142