Garcia Papers in Open Access Journals

[JCDowdy]

Dr Dyer sent me some links to share with you all to some recent papers with his colleague Dr Jair Garcia who recently completed his PhD on using linearized sensors to make quantitative measurements with both UV and RGB cameras. He says he and Dr Garcia have a few more manuscripts in preparation and they shall be touring UltravioletPhotographt.com as time permits.

 

Garcia JE, Greentree AD, Shrestha M, Dorin A, Dyer AG (2014) Flower Colours through the Lens: Quantitative Measurement with Visible and Ultraviolet Digital Photography. PLoS ONE 9(5): e96646. doi:10.1371/journal.pone.0096646

http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0096646

Garcia JE, Rohr D, Dyer AG. (2013) Trade-off between camouflage and sexual dimorphism revealed by UV digital imaging: the case of Australian Mallee Dragons (Ctenophorus fordi). Journal of Experimental Biology.

doi: 10.1242/jeb.094045.

http://jeb.biologists.org/content/216/22/4290.short

Garcia JE, Dyer AG, Greentree AD, Spring G, Wilksch PA (2013) Linearisation of RGB Camera Responses for Quantitative Image Analysis of Visible and UV Photography: A Comparison of Two Techniques. PLoS ONE 8(11): e79534. doi:10.1371/journal.pone.0079534

http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0079534#pone-0079534-g004

Interesting stuff, I am looking forward to seeing more of their work.

- John

[DaveO]

For what it's worth (probably not much) here's the link to my shots of Goodenia ovata which is one of the flowers used in the first paper

 

http://www.ultravioletphotography.com/content/index.php?/topic/680-goodenia-ovata-hop-goodenia/

 

Dave

[Andrea B.]

Thank you for these links, John.

We will be perusing these papers for sure.

[JCDowdy]

... here's the link to my shots of Goodenia ovata which is one of the flowers used in the first paper....

 

Dave,

 

Nice to see a good comparison of the techniques. In the 2014 PLOS ONE paper, Figure 4.b)box II, their linearized UV-intensity pseudocolor renders significantly more contrast to your false color UV-yellow on UV-yellow nectar tracks.

 

I suspect that there may just as likely be cases where contrasting false colors with similar reflectance intensity would be rendered more clearly in the "ultravioletphotography.com" reproducible false color palette.

 

Google Maps says Melbourne to Maldon is only ~140 km, perhaps you guys might collaborate!

 

- John

[Andrea B.]

FIRST PROBLEM in the Linked Papers: You cannot properly convert a Fuji S3-UVIR raw RAF file in Adobe ACR.

Do you know why?

 

 

(Hint: This camera has two types of pixels.)

[Andrea B.]

SECOND PROBLEM in the Linked Papers: The Fuji S3-UVIR won't record below 350nm as stated by Fuji. So we don't understand why the linearized UV response in Figure 3 of the first paper shows a range of 320-395nm ?? Were the authors just giving the range of the Baader-U there? Or do they think they have recorded below 350nm with the Fuji S3-UVIR ? I need to re-read this part again.

 

Bjørn R. tells me that he tried to use the Fuji S3-UVIR with a 340nm BP filter and couldn't really get an image.

[Andrea B.]

THIRDLY: Can't you just pull the linear data from the raw RAF file using Dcraw ? I don't know why this is considered a problem in the papers requiring Mathlab and special algorithms and Bezier curves ?? And you don't need to squash the data further by going to an 8-bit save as they do. You can get 16-bit linear data easily in Dcraw.

 

Oh well. Maybe I'm missing something here.

 

ADDED LATER: Dcraw produces from a 16-megapixel RAF a weird 16-megapixel TIFF output file tilted at a 45-degree angle. The only reason the TIFF 16-megapixels is that there are 6-megapixels worth of black pixels around the actual photo. Boo, Dcraw !!! You have let me down.

[DaveO]

I think I'm missing something as well.

If only the 'red' UV-sensitive channel of a Fuiji S3 UVIR camera is used HOW do they get the false reddish colour from the NIST white target????

 

Here's what my Eremophila maculata looked like in false color UV

 

[Alex H]

THIRDLY: Can't you just pull the linear data from the raw RAF file using Dcraw ? I don't know why this is considered a problem in the papers requiring Mathlab and special algorithms and Bezier curves ?? And you don't need to squash the data further by going to an 8-bit save as they do. You can get 16-bit linear data easily in Dcraw.

 

Oh well. Maybe I'm missing something here.

 

I think the problem here is that no-one really knows how much processing is done to the data before it is saved as RAW file.

[nfoto]

They stated ACR 6.3 and CS5 were used to extract data. It is well known ACR does not understand the special S+R pixel structure for the Fuji S3. In fact, almost no software I'm aware of does this. PhotoNinja does not although it runs the Fuji RAFs. Fuji's own clunky and awkward software does as per the maker, but what you get out of the file is pretty much obscure. A free program S7RAW handles Fuji files and indicates by way of its multitude of settings how complex the conversion of the S+R structure really is. However, S7RAW hasn't seen updates for many years and is a dead end.

 

I purchased the Fuji S3 Pro UV/IR Limited Edition many years ago and still use it occasionally, but not for UV. This because its UV response really isn't much to write home about compared to the better alternatives. The spec sheet accompanying the camera clearly states spectral response 350-1000 nm and actual shooting bears this out. Even with my Bron studio flashes (uncoated Xenon tubes) running at full power, there isn't much if anything to be had through narrow-band filters below 350 nm, so I consider Fuji's specification to be on the mark. Fuji launched the IS (derived from D200 if memory serves) later for forenisc use and here they stated a response from 380 to 1000 nm. I think the S+R sensor of Fuji is to blame for the reduced UV performance. A pity the authors weren't aware of these limitations before they put a lot of efforts into massaging RAFs from this camera.

[Andrea B.]

Alex, I grant you, we are likely getting what the cam manufacturers consider "raw" after readout, prior to conversion. Even more reason go be cautious about analysis.

 

I note that Raw Digger has a research edition and wonder if it could be helpful to these guys. Dcraw to extract the uncurved foto and Rawdigger to analyze the data.

Research Edition is meant for camera/sensor evaluation and statistical analysis of raw data. It is suitable for a wide range of tasks, starting from personal use and preparing reviews and all the way to programming RAW data processing. Research Edition adds TIFF export, multiple selected areas via samples, tables of sampled data; statistics and histograms over multiple samples, exporting sampled data as CSV and CGATS.

 

I would also like to see some actual data on the Bayer filter dyes. That must exist somewhere.

 

Perhaps we could suggest to the paper authors to repeat some of the experiments with modified D600 or Sony Nexi cams ???

 

I also once had a Fuji S3-UVIR. I would say that 350nm is a generous estimate.

 

It is always easy to find flaws. Better to try to make suggestions for a better experiment.

Please add to this list of Suggestions:

• Use low-noise camera known to record below 350nm - example Nikon D600, Sony Nex6.
  
• Use better lighting - example: Broncolor UV.
  
• Examining raw data and extract foto with photographer's tools - example Rawdigger and Dcraw and [ADDED] the old Hyper software Fuji used to offer.
  
• Proper 14-bit conversion of photograph and saving as 16-bit TIFF for work in/on 64-bit editor/machine.
  
• Consideration/analysis of Bayer dyes.
  
• Perform some tests with narrow bandpass UV-pass filters.
  

[Alex H]

Alex, I grant you, we are likely getting what the cam manufacturers consider "raw" after readout, prior to conversion.

 

Andrea, I am not sure I understand you.

 

All I was saying is that we do not know how much processing of signal is done by the camera before the file is saved as RAW (if any at all?). No converter can "revert" this process, if I understand it correctly.

Thus, I agree with authors on the use of pre-measured illumination for the input, and the use of special algorithms to extract linear data from the "raw" file.

[Andrea B.]

Saying basically the same thing, but not too well. :D

However, I'm not sure that linear data can't be extracted more easily however.

Especially for non-RAF files.

 

**********

 

We just experimented with converting a Fuji Pro S3-UVIR in several converters: Photivo, Photo Ninja, Raw Photo Processer (RPP), Iridient Developer, Darktable and Dcraw. Unfortunately NONE of them were capable of properly converting the 12-megapixel RAF files by making use of both the S pixels and the R pixels to produce a 12-megapixel TIFF output file. All produced 6-megapixel output files. (One converters claims to "mix" or blend the two types of pixels but it is not evident what benefit there is to that in the subsequent 6-megapixel output - at least for the file I was working with.)

 

Bjørn then dug out his ancient and creaking copy of Hyper-Utility software originally supplied by Fuji with the S3-Pro cameras. Upon comparing a proper 12-megapixel file produced by Hyper-U with one of the 6-megapixel output files from another converter, it was easy to see that detail is lost when half the Fuji pixels are thrown away in the conversion step (or improperly "blended").

I'm hoping he will post some sample images later when he gets the time.

[JCDowdy]

If only the 'red' UV-sensitive channel of a Fuiji S3 UVIR camera is used HOW do they get the false reddish colour from the NIST white target????

 

Dave,

That is not false color UV as you posted. As I read it that is a pseudocolor LUT of linearized intensity from the essentially monochrome Fuiji red channel, the adjacent vertical bar is the scale.

John

[JCDowdy]

Andrea & Bjørn,

 

Are you still on expedition in Norway?

If so please accept my apologies for causing such a distraction - :D

 

I hope to see your critique answered by the authors if and when they join the group.

I must confess being surprised by the technical limitations of the Fuiji UVIR camera you describe, having not heard much praise for it here I guess I should not be.

 

ADDED: Just remembered I found this some time ago when scouring Jenoptik website:

http://www.jenoptik-inc.com/coastalopt-standard-lenses/uv-vis-nir-60mm-slr-lens-mainmenu-155/314-recommendations-for-uv-imaging-with-a-digital-slr.html

 

- John