What is the real response of Panasonic Lumix G3? False colors vs. wavelengths. Spectroscopy.

[DaveO]

Here's a ref for Colour Wheel http://en.wikipedia.org/wiki/Color_wheel#mediaviewer/File:RBG_color_wheel.svg

 

which puts the combinations of RGB together and names them pretty well.

 

Dave

[colinbm]

Colour Wheels are unnatural....colour is not circular but linear.... there is no magenta in the rainbow.....

http://www.biotele.com/magenta.html

Col

[JCDowdy]

Aren't most of our UV cameras only really responding in two channels?

 

No. My Raw Digger histograms show response in all three channels. Usually the R channel has more.

 

Yeah, I did not state that correctly. What I am trying to articulate is the perception that our out of band UV response seems to not be fully trichromic, at least with the typical UV neutral WB and false color palette response. We get these nice looking blue-gray to yellow images with UV neutral WB. I don't want to get tangled up in descriptive color terms, but the idea is we with neutral UV WB we are not seeing a full trichromic false color palette. I am frustrated for lack of more correct nomenclature to express this. Lets say for some reason, light or luck of reflectance, there is a shorter wavelength than normal signal. Suddenly an abnormally vivid green jumps into the mix. Is this a WB error or artifact? Is it because when the system is WBalanced in the absence of, or only weak presence of, a spectral component the whole rendering of that region is skewed or abnormally amplified? Please don't misunderstand, this is more of an observation of my perspective than a specific question, lacking obviously in technical underpinning, only food for thought.

[Andrea B.]

I understand what you are trying to express.

 

Remember that most of the UV work you see here from Bjørn and me that has been "standardized" into the blue/yellow palette was made with a Baader U filter. That filter is probably not wide enough to capture a fuller UV false colour palette because it only has about a 60nm wide half-bandpass. Then too, the amount of sunlight UV tapers strongly as we go from 400nm down to 300nm, so we are likely capturing much more UV between 350-385 than we are between 315-350nm under a Baader U even though it peaks around 350-360nm.

 

When the experimenters are making their false colour charts they are making readings over the entire 300-400nm range (and sometimes a little below 300nm), so it is not unexpected that they would generate more false colours.

 

That's my best guess right now. But I will think about it some more. ;)

(And also go check some facts.)

[JCDowdy]

Good, then I am at least intuitively understanding the bigger "picture" as it were. I always thought that the bandpass of the Baader U was actually narrow, to me anyway, compared with the ~380-770nm normal visible range. Also, given my line of work, I am certainly familiar with the overlap of the solar UV spectrum with that filter. One aspect that for me remains is that the KDS false UV color palatte, with Sascha's recent seeming confirmation, as well as Enrico's work in the area, is that I thought these false color palettes were made with the Badder U and not unfiltered camera response. Am I wrong on that point?

[enricosavazzi]

[...] I thought these false color palettes were made with the Badder U and not unfiltered camera response. Am I wrong on that point?

The only reliable way to compile a UV false-color palette with a full-spectrum camera is by using narrow-band filters, which are generally defined as having a 10 nm bandwidth (at 50% attenuation relative to the max transmission peak). Once you have compiled a palette, you can then shoot with a UV-pass filter with a broader (but not so broad that it includes VIS or UVB) bandwidth, like the Baader U, and from this image you choose an area of false color and find the closest match in the palette. A good match usually (but not always) indicates the same prevalent wavelength + or - about 10 nm. Lack of a good match (or sometimes a good match) may be due to a mixing of wavelengths reflected by the subject. To be sure, shoot with the same narrow-band filter that is supposed to produce the observed false-color. If the same area gives the same false-color also with the narrow-band filter, you have a positive identification of the wavelength interval.

 

Alternatively, you can use a monochromator to produce the false-color palette.

[Andrea B.]

Monochromator, no lens on camera, no filter on camera, camera mirror up, aim light inside chamber inside chamber, shoot. More or less.

Lots to do there with blocking ambient light. Work in darkness??

Must strongly filter source illumination to block the stray rays.

 

Do I have the gist of it? :D

 

**********

 

I've looked into narrow-band UV-pass filters more than once, but they are expensive and small in diameter. I would not be using them for creating any false colour palette. I would enjoy them simply as cool experimental tools to discover what's "out there" in fauna & flora, photographically speaking.

 

**********

 

Enrico, I surely do hope you continue to work on the false colour palette problem. There are so many interesting variables to nail down.

 

**********

 

I do want to continue to define and describe the experimental protocol for a false colour palette. Then if a double-grated monochromator/spectrometer ever falls from the sky, I will be ready !!! ;)

[enricosavazzi]

Monochromator, no lens on camera, no filter on camera, camera mirror up, aim light inside chamber inside chamber, shoot. More or less.

Lots to do there with blocking ambient light. Work in darkness??

Must strongly filter source illumination to block the stray rays.

 

Do I have the gist of it? ;)

That is one way of doing it, and that is the way Sascha did it. Of course you must exclude other radiation. Cheap monochromators often leak substantial amounts of radiation at other wavelengths, which is not good.

I've looked into narrow-band UV-pass filters more than once, but they are expensive and small in diameter. I would not be using them for creating any false colour palette. I would enjoy them simply as cool experimental tools to discover what's "out there" in fauna & flora, photographically speaking.

A couple of eBay sellers offer very small narrow-band filters that can be quite cheap and can be used for creating a false color palette (not the way of Seeve Smeed's filter strip, but by taking one picture with each filter, with the filtered centered on the optical axis of the lens or sensor, because these are interference or hybrid filters). For imaging, unless nothing else is available, the minimum filter diameter for reasonable use is about 20 mm. Some of the narrow-band filters, like Thorlabs' (unless they changed during the last couple of years) are thick multilayer sandwiches that may give now (edit: now->low) contrast and relatively low resolution images. Single-layer filters are preferable. I chanced on a 325/10 Omega filter that gives good high-contrast images.

 

I think narrow-band filters usable for imaging are useful, because one can use them for testing as well as imaging. We must not forget, however, that the spectra of narrow-band filters have broad low-transmission heads and tails, so if the sensor's sensitivity drops off sharply below 350 nm, even a 325/10 filter may actually give an image recorded mostly at 340-350 nm, so no free lunch yet.

[colinbm]

Back in the first post of this thread, & this link to a UV false color palette

http://m8.i.pbase.co..._UVcolors_c.jpg

I am wondering if this is a UV Induced Fluorescence result of the Bayer CFA (which is commonly acknowledged), but also remembering that the camera's output is an interperpolation of four Bayer CFA dyes, to give one mixed colour. That is there are a group of four coloured pixels (R, G x 2 & ;) , that have each their own fluorescent properties in VU light, this group of four pixels are then combined (interpolated) to one colour value, which is what is seen in the above image.

Col

 

PS, If only, we could see each, Bayer CFA, RGB dyes, performance under UV lighting........

[Andrea B.]

Back to Enrico's post #57.

 

Certainly it is true that

if we made a UV foto with the broadband Baader-U

and if we matched the colour of a small area in the resultant foto to a wavelength in the false colour chart,

we could then confirm that this particular false colour to UV wavelength correlation was correct

by reshooting that small area using a narrowband filter around that wavelength.

 

But what would be the point of doing things that way?

Why not just directly shoot the subject with the narrowband UV-pass filter in the first place?

 

If you record a bright area (UV reflectivity) under a narrowband UV-pass filter,

then you automatically know you have a reflected UV wavelength within that narrowband

and it is not necessary to deal with false colour at all. ;)

 

In other words, we do not really need a False Colour Chart,

we need instead a good set of steeply-shouldered, toeless, narrowband UV pass filters.

 

Any comments on this proposition ??

[JCDowdy]

That is the classic way of doing scientific multispectral imaging with a monochrome camera back and filters. The renowned lens designer Brian Caldwell has designed a UV-Vis-NIR super lens for this system. The big difference of course is that it is not an out of band CFA transmittance but a well defined spectral band. Astronomy filters from Baader abound in many colors for this very purpose. With a suitable set of filters, non existent as far as I know, in the UV one could recombine UV bands into a pseudocolor UV palette but then you have come full circle and must decide what color to put where! ;)

[enricosavazzi]

But what would be the point of doing things that way?

Why not just directly shoot the subject with the narrowband UV-pass filter in the first place?

 

If you record a bright area (UV reflectivity) under a narrowband UV-pass filter,

then you automatically know you have a reflected UV wavelength within that narrowband

and it is not necessary to deal with false colour at all. ;)

 

In other words, we do not really need a False Colour Chart,

we need instead a good set of steeply-shouldered, toeless, narrowband UV pass filters.

 

Any comments on this proposition ??

My reason for shooting with a broadband filter like the Baader U and then trying to identify wavelengths from false color is practical. I only need to take one picture to get a rough idea. If I notice something unusual, I can then go in with narrowband filters to get a better idea and to resolve distinct wavelengths that create overlapping false color. It would be more reliable to shoot five or ten separate images with different narrowband filters, but the thought of doing this handheld in the field on a windswept flower while manually changing five or ten filters that cost $300 each is not particularly appetizing to me. :D

[OlDoinyo]

I have found this thread difficult to follow, at times. It seemed to have started as an attempt to map the spectral response of a camera, and then somehow white-balance seems to have crept into the topic. I would have approached the initial task as follows:

 

1) for each channel separately at each wavelength at a given sensor gain (ISO), measure the exposure time needed to produce a given standard pixel brightness in the target area (e.g. 32767 in a 16-bit system)

 

2) normalize by dividing by a multiple of the (presumably known) spectral intensity of the reference light source at each wavelength, convolved with the monochromator transmission curve;

 

3) plot the three graphs versus wavelength.

 

With this information in hand, it should be possible to calculate whatever white-balance correction might be needed at each wavelength.

 

Yes, monochromators can be leaky, some worse than others. Back them up with filtration--or use a tunable dye laser for a light source rather than a deuterium lamp.

[Andrea B.]

Enrico, that is certainly all true about convenience and cost. :D

 

It would not solve the problem of a double-humped UV reflectivity between 300-400nm for some flower.

If there is such a flower. ;)

 

**********

 

Clark, we've all been hacking at this thing for a long time now. So the thread has naturally wandered into peripheral topics associated with UV wavelength to false colour correlation.

I have to think about your method for awhile.

[Andrea B.]

There was a discussion on Dpreview (of all places!!) which touches upon some of the observations in this thread.

 

TITLE: Some observations on handling greens and violet in raw conversion.

LINK: http://www.dpreview.com/forums/thread/3782319

 

It is observed in this discussion that pure spectral violet has a "purple" colour but so does a mixture of red & blue light taken from the opposite sides of the spectrum.

 

Clearly when we record Visible purple in our cameras - we do not really know whether we were recording the spectral violet or the red&blue mix.

 

Now translate that observation to the UV situation - a pure spectral UV wavelength versus a mixture of UV wavelengths - both giving the same false colour. How do you know which wavelength you have recorded?

 

Observations are also made about the "twist" in colour spaces. This could definitely affect any attempt to make a UV-Wavelength to UV-False-Colour correlation. (I need to add that to the Variables above. I had forgotten about Twist.)

[Andrea B.]

Digging up old bones.....sorry 'bout that. But I felt like I needed to add something to this controversial* topic about the fact that the false colors from my converted Panasonic S1R were a bit different from the (so-called) standard false colors which we have seen for the last 10 years or so. I put this update in the first post.

 

*This topic was controversial in 2015. Arguments were started. Hackles were raised. Fur flew right and left!

Maybe we have all calmed down since then. 

 

EDITOR'S NOTE: UPDATE 19 MAY 2022.  My full spectrum converted Panasonic S1R produces false colours in white balanced, reflected UV photographs which are slightly different from what we have seen so far with full spectrum conversions used with broadband UV-pass filters. There seems to be more dark cyan-green or green. We are thinking that this indicates a slightly different Bayer dye somewhere in the Bayer filter.

Anyway, I think I therefore need to add to the discussion the fact that false-color-to-wavelength charts are only good for the specific camera from which they are made in case that has not been obvious. Or, for those of you who think that false-color-to-wavelength charts are not good for general use, now you have another reason why. They may vary by camera.

[dabateman]

Too bad Sascha never came back and only had 7 posts.

I see the same color pattern per wavelength as Sasha for my Olympus Em5mk2 and Em1mk1 cameras.