Color artifacts in a reflected UV botanical photo

[Andrea B.]

Added Later:  My summary has become this:  why is there color noise in the UV photo but not in the Visible or UV+Blue+Green photos?

---

Added Later:  I attempted an explanation here - LINK.

---

 

This is an unresized crop from a D610 + UV-Nikkor + BaaderU + SB140 photo of Chamaebatiara millefolium photographed indoors against a black velvet background. The file has been converted and white balanced only. No sharpening or detail enhancement has been applied.

 

The flower buds, stems and small leaves are very, very hairy. The areas with these small hairs show lots of color artifacts. I don't think these are due to iridescense. It seems rather that the complexity of the hairy areas causes some kind of moiré like effect. I'm not sure what is the correct terminology for this effect.

 

This will click-up to approximately 1600 x 1800 pixels in an enlarged browser. Then you can see the color noise. But next after this photo is a 3X enlargement which clearly shows the color noise.

 

 

This screen shot was taken from a 3X enlargement in Photo Mechanic. There are cyan, pink, blue, brown and green areas.

 

 

 

Here is that same area after Noise Ninja color reduction set to the default 50.

Not quite every color is gone, but things look a bit less color-noisy.

 

 

Finally here is Noise Ninja at the maximum 100 for color noise reduction.

There are still some colorful bits, but the hairy areas are much more neutral.

 

 

 

Here's the first thing:  I think I like seeing the color noise due to the complexity of the hairy areas. So I'm undecided about whether to de-noise a photo like this or not.

 

Here's the second thing:  The UV light passes through the Bayer filter and is primarily recorded in the red channel. The white balance step produces typically some combination of false blue, false yellow and grey/black/white tones. So where does this color noise come from? It must be a result of demosaicing?

 

 

This is an enlargement in Raw Digger showing the file before any white balance is applied.

You can see some of the color noise in this raw composite.

[Andrea B.]

But then there is this - made with the U330 x 2.0mm + S8612 x 2.0mm to give a UV+Blue+Green file. After demosaic and white balance only, the enlarged hairy bud area shows NO color artifacts other than what appears to be a couple of hot pixels and a few bright pixels. What is it about the pure UV version above that produces the color noise in the hairy areas?

[Andrea B.]

To be boringly complete....

This is an enlargement of the demosaiced, white-balanced Visible version of the photo. There is no color noise in the complex hairy areas.

(Note: The Photo Ninja exposure slider was also applied on this file to bring down some overly bright areas.)

 

[Andrea B.]

My Summary

 

There is a lot of color noise in the complex hairy areas of this plant in the reflected UV photo. This color noise does not appear in the Visible version or in the UV_Blue_Green version. I don't have an explanation for this.

 

I should look at possible color noise effects from the dichroic BaaderU?

[Fandyus]

I use Tangsinuo glass and I have the same issue. Always wondered what was the cause.

[Andy Perrin]

Andrea, try processing it with one of the other Raw processors like Darktable. I definitely suspect PN has a poorly designed demosaicing module.

[Andrea B.]

Here is an unresized excerpt from the file as demosaiced and white balanced in Capture NX2, Nikon's native converter. (As a side note, NX2 does not apply the same "gamma" (curve adjustment) as Photo Ninja.

 

 

Here is an enlargement of the hairy buds from the NX2 version.

 

[Andrea B.]

 

Here are some demosaics from Raw Therapee, which would not make a proper white balance. We can try to look at the noise anyway. Raw Therapee has 10 demosaic options.

 

Demosaic:  Amaze

Clearly has color noise.

 

 

Demosaic: Lmmse

This demosaic option is said to be good for high ISO to aid in noise reduction without producing maze patterns. This option is the best of the 4 shown here. But there are still too many extra colors.

 

 

Demosaic:  Vng4

Has color noise.

 

 

Demosaic: Ahd

Clearly this is not a good choice! It has chunks and also color noise.

 

[enricosavazzi]

Just a possibility, but the color artifacts could be a result of in-camera image sharpening and demosaicking originally designed for the image produced by a camera with a physical anti-aliasing filter. If the anti-aliasing filter is removed during conversion, the image is sharper than expected and color moiré effects are more likely to appear. The demosaicking algorithm instead may still be optimized for the original image resolution with the anti-aliasing filter still in place, and overcompensates as a result.

 

The small hairs may be at the edge of the maximum sensor resolution, which makes the problem more visible. False UV color may also play a part in the problem and not play well with the demosaicking algorithm, since the Bayer color filters are designed for VIS and work differently than expected in UV.

 

Largely monochromatic images like the one taken with the U330 may lessen the problem. Perhaps resolution in this image is also overall less than with the Baader U.

[Andrea B.]

Good points to think about.

[Andy Perrin]

Just a possibility, but the color artifacts could be a result of in-camera image sharpening and demosaicking originally designed for the image produced by a camera with a physical anti-aliasing filter. 
 

I don’t believe this should be the case for images processed from RAW. In-camera sharpening only applies for JPEGs shot in the camera. Andrea is shooting RAWs and demosaicing afterwards. 
 

The small hairs may be at the edge of the maximum sensor resolution, which makes the problem more visible. False UV color may also play a part in the problem and not play well with the demosaicking algorithm, since the Bayer color filters aredesigned for VIS and work differently than expected in UV.

 

This seems plausible though, even with RAWs. 
——

 

Andrea, remember this cinnamon stick? I wonder if all those colors are real or if there is a demosaicing issue at play…

 

https://www.ultravioletphotography.com/content/index.php?/topic/2471-ir-visual-uv-ug1-animation/#comment-18130

 

[dabateman]

Andrea that looks like aliasing  artifacts at the pixel level,  that are being discussed over on the Fujifilm 50S vs 100S thread here:

https://www.dpreview.com/forums/thread/4632624?page=4

 

If you are at the pixel level,  for those hairs and not getting color blur, that might be the case.

[Andrea B.]

OK, I've been rummaging around the internet. I might have the beginnings of an explanation.

I don't quite know the proper technical language for this explanation, so please bear with me.

 

Explanation

 

I think that false false-colors⁽¹⁾ are present in my reflected UV photo because details of the fine hairs on the stems and buds of the plant are not correctly "interpreted" by the sensor & Bayer array when shooting in UV. If the plant hair is smaller than one of the Bayer squares, then the square "sees" both the hair and its background. So two reflected wavelengths are "aimed at" the Bayer square which will record the blended reflection instead of the reflection off the hair alone.⁽²⁾ For larger areas of this plant, such as the petals, the details being recorded are much larger than any single⁽⁴⁾ one of the Bayer squares so there is no blending of reflected wavelengths. 

 

These tiny color anomalies do not happen in the Visible (or UV+Blue+Green) photo above because Visible tiny details are blurred by diffraction⁽³⁾ which sets in earlier than it does at the same aperture in a reflected UV photo. The details become blurred enough to fill the Bayer square entirely with no confusion from any background reflections. 

 

Comments

I don't know what the proper terminology is for these false false-colors in a reflected UV photograph. It isn't exactly a moiré interference pattern. It's more like a small detail resolution artifact.  

 

The sharpness of the lens must also plays a role in this kind of artifact occurence. Less sharp lenses might not have as much UV false false-color.

 

I am totally wiling to accept that I got this completely wrong. It wouldn't be the first time.

Sometimes we get misled⁽⁵⁾ by the internet. Sometimes we misle ourselves.

 

Footnotes

 

(1) Perhaps I should use the phrase "anomalous false-colors" rather than "false false-colors".

 

(2) Of course the blended wavelengths might also be "rejected" by a Bayer square, but I'm trying to not be tedious.

 

(3) We all already know about diffraction differences in UV, Vis and IR photos, but it needs to be mentioned as part of the explanation of false false-colors.

 

(4) I made this explanation with reference to only a single Bayer square. It is possible that we need to consider how fine details and their background are "interpreted" by a typical block of Bayer squares (such as 1 B, 1 R and 2 Gs) when the detail is smaller than such a block.

 

(5) American Slang. misled = pronounced MI-zuld. misle = pronounced MI-zul. The 'i' is sounded like 'eye'. Politicians are very good at MI-zul-ing people.

[photoni]

probably in my life I was influenced by the film Blow-Up by Antonioni,

in '66 they showed things impossible to obtain from a magnification of a Nikon F 24x36 mm

Andrea ... you now expect more from a D610 :)
I believe the difference in the UV file is because it is underexposed.

somehow the interpolation of the three channels to soften the passages doesn't work.
Try to change soft and not indicate the type of DSRL.

Antonio

[Andrea B.]

I saw that film !! Sex, drugs & rock-n-roll along with cameras and murder (or not)!

Perhaps there will be a remake using a digital camera.

But a remake would never capture Antonioni's artistry.

 

I'll check the exposure of the photo for under- and over-exposed areas.