Question to Raw Digger with Response

[Andrea B.]

I thought I'd list this separately from the topic in which it arose.

 

Here are my questions to the developers at Raw Digger.

(You can skip to the bullet points.)

 

Ultraviolet photographers have made a natural mapping from reflected UV wavelengths to white-balanced false colours produced by the Bayer-filtered digital sensor (using a camera having its internal UVIR blocking removed) under a UV-pass filter.

 

For example, yellow false colours are produced by shooting in sunlight through a narrowband UV-pass filter peaking around 340nm while violet/blue false colours are produced by a filter having a 380nm peak. This is also seen under broadband UV-pass filters. The elephant-in-the-room is metamerism of course. It is entirely possible that some combination of reflected UV wavelengths or illumination differences might also produce a false yellow. It is always a many-to-one mapping, in theory. [The smaller elephant in the room is that to get from wavelength to RGB, we are making a **lot** of approximations.]

 

However, it has been suggested that white-balanced false colour UV photograph of say, a flower, might be used to determine *in reverse* the reflected wavelength (singular) from the petal producing the false colour. This reverse map has proven somewhat valid tor those cases where we have spectrographic measurements to back it up. But given that any digital colour, whether visible or false UV, is dependent upon such a large number of variables including illumination and Bayer dyes in addition to white-balance algorithms, this reverse mapping can only be applied with great care.

 

For the record, with a converted camera, a professional UV lens like the 105/4.5 UV-Nikkor, and a broadband UV-pass filter, we are able to record between the 300-400nm range. We probably get 60nm (?) width of good signal. In the visible range a 60nm width would be enough to produce 3 close but different colours depending on the peak, so I assume this holds somewhat true for the UV band to produce at least 2-3 false colours from good signal.

 

===>>>>> Recently I have been looking at my UV files in Raw Digger to try to determine whether the raw false colours might tell us something more useful then the white-balanced false colours. I note that there do seem to be 3 raw false colours which I commonly see. Red is the primary channel in which UV is recorded so I regularly see Pink/Red, Orange, and Purple/Magenta in the raw colours as follows: Pink/Red [big R, mostly equal but small G and B]. Orange [big R with G from .3 to .5 the strength of R, small B]. And Purple/Magentas [big R, B larger than G]. Some filters pass a bit of violet near 400nm to produce more B in the false colour.

 

The production of these false colours is remarkably similar across platforms having Bayer filters. After white balance we reliably get a UV false colour palette of blues, yellows and grey/black/white. Typically only false blue or false yellow appears against the neutral tones. But sometimes we get both blues and yellows.

Example: http://www.ultraviol...-false-colours/

 

So, my questions:

• Does viewing the raw UV false colours in Raw Digger strip away the variation which might be induced by different white-balance tools in different converter apps? (Note: I use a Labsphere spectralon reflective target for white balance in UV.)
  

• Does a white-balance step eliminate the effect of the Bayer dyes and thus indicate that viewing white-balanced false colours is preferable over viewing raw false colours when we want to attempt a reverse mapping back to wavelength? (Still ignoring that elephant.)
  

• Are there any "hidden" factors such as camera profiles or whatnot in the Raw Digger conversion which I need to know about when examining raw false colours? I don't think so, but prudence requires I formally ask. "-)
  

 

 

And here is the response from Iliah Borg at Raw Digger.

Thanks, Iliah !!

(The bolding is mine.)

 

Technically, back-side illuminated sensor gives some response up to 200nm. I prefer those BSI sensors when I work out of band (makes for nice welding arc shots). Bayer CFA reduces this to something around 340nm. You can look at Foveon-based cameras with stripped protective-filtering cover glass to get rid of CFA effects, however you will be getting the image mostly in the first layer, amounting to b/w. Sometimes, when several shots through different narrow band filters are possible, it is a good way to get "colour".

 

Converters apply colour transforms (colour profiles) and white balance (also in fact a colour transform). This is a strong factor, affecting the colour appearance. With some converters one can use a custom profile which is a simple eye matrix, doing nothing in terms of mixing colours between channels.

 

In RawDigger, no white balance is applied when the mode is set to Raw composite, and of course no colour transforms at all are applied when one is viewing raw channels in Raw channel mode. Those channels can be exported from RawDigger as separate TIFF files for future processing in some standard application like Photoshop, and the red channel may be used as L channel when synthesizing the image from separate channels.

 

No, white balance can't completely eliminate the effect of Bayer filters. Different cameras still use different CFA systems, which amounts to different spectral responses.

 

I hope I was able to answer at least some of your questions.

 

Please give my best to Bjørn.

 

Best regards,

Iliah Borg

LibRaw, LLC

www.libraw.org

www.rawdigger.com

www.fastrawviewer.com

[Andrea B.]

So UV false colour palettes work only on a per camera basis because of the Bayer filters. (In case anyone had not yet observed that.) If you want to try to reverse map false colour to wavelength, you must correlate wavelength-to-response for your individual camera+lens+filter combo in use.

 

A standardized (Bayer filtered) colour palette for UV false colour -- like we use here in the botanical section -- will never be perfect. There will be variations.

(It still amazes me that we get "similar" colours across such a wide variety of gear.)