Jump to content
UltravioletPhotography

Interesting WB Result & Flower Structure


Andrea B.

Recommended Posts

I was working up the photo of the abaxial side of this Desert Marigold to possibly add to the topic. The white balance result was *very* interesting.

 

Gear: Nikon D610/fullSpec + UV-Nikkor 105/4.5 + BaaderU UV-pass Filter + SB14 UV-flash Mod

 

Here is the back side of the Desert Marigold as converted in Photo Ninja and with all the usual editing except for white balance.

Click up to get 1200 pixel width.

baileyaMultiradiataAbaxial_uvBaad_sb14_20200523aveVistaGrande_19883pn01X.jpg

 

 

The flower was photographed against a Spectralon square which makes WB easy. I simply dragged the PN white balance dropper around some of the Spectralon area.

(Yes, there is grunge on the Spectralon. Kindly ignore.)

baileyaMultiradiataAbaxial_uvBaad_sb14_20200523aveVistaGrande_19883pnWbX.jpg

 

 

But what an unusual WB result !! I sampled the colors and found pinkish-browns and light cyans. These colors are not "supposed" to occur in a properly white-balanced UV photo. Yet here they are. This is an unresized crop.

baileyaMultiradiataAbaxial_uvBaad_sb14_20200523aveVistaGrande_19883pnColorX.jpg

 

 

Here's a screen shot blow up to make it easier to see the weird colors.

blowup.jpg

 

 

Is this result because of color noise? In ample light the D610 is just too good at ISO-200 to produce much color noise. So I am thinking that these unusual pinks/cyans have something to do with the flower ray (petal) structure? Flower rays and petals can transmit light as well as reflect light. Transmitted light hits internal structures or interacts with surface cells and the result is iridescence or other odd effects. Is there some moiré here? Maybe this is some kind of demosaicing artifact? Moiré, maybe?

 

But this photo was under a BaaderU. How do false pinkish-brown and false light cyan fit into the usual blue/yellow/gray/black/white palette which we see after white balancing? Maybe this is color noise. But shouldn't color noise under a BaaderU be not-pink and not-cyan?

 

What do you think?? :cool:


 

I converted the photo in old Capture NX2 and got the following result. It looks rather different but still has some odd colours. Interestingly, NX2 had trouble white balancing this. It left some cyan highlights on the Spectralon. That was strange.

This is a blowup of an unresized crop so that the colors will be more easily seen. I think that NX2 does not apply as much saturation as does Photo Ninja.

screenie_nx.jpg

Link to comment
My guess is color noise. I know we can't derive wavelengths from false colors, and this has been discussed a lot. But, usually, upper UV (~380-400 nm)is blue-lavender, UV around 360 nm is yellow and around 340 nm is green. There's no room for cyan and pink. Also those colors are complementary and produce gray.
Link to comment
Andy Perrin
I agree. Color noise is unrelated to the light source. You can prove it by taking a photo with the lens cap on. No light, but still plenty of color noise. Also, it's in addition to any signal generated by the light source - you can have both signal and noise in the same pixel.
Link to comment

I know we can't derive wavelengths from false colors

 

It is not a simple black or white thing whether you can't or can.

Of course you can make reasonable assumptions about what wavelength is being reflected. But you cannot be sure.

 

It does hugely bother me that these wavelength assumptions are made on white balanced photos. There is a lot of trauma caused to a file when we made these extreme UV white balance efforts.


 

Back to the weird photo above.

There isn't any color noise in the raw photo. It only appears after the white balance step.

Link to comment

Back to the weird photo above.

There isn't any color noise in the raw photo. It only appears after the white balance step.

As you said, the white balance is extreme (I am honestly surprised most cameras can do that). That can amplify color noise (I don't know how much, perhaps quite a lot).

 

Trying the same experiment on other surfaces with similar characteristics can tell us more.

Link to comment

Here is the front of the flower. Just pollen grains all over in blue or green. But front and back were photographed in the same session with the same gear.

The front is a lot darker than the back in UV.

 

Click on this to get big version.

baileyaMultiradiata_uvBaad_sb14_20200523aveVistaGrande_19839pnNewCrop.jpg

Link to comment

So... here you have pollen in blue and green. Those are colors that can occur in UV photos, but green is unlikely to be seen. Blue is more common, but I don’t think it comes from the pollen’s reflectance properties. If you look closely, there are a ton of little colored dots. To me, that looks like noise.

 

We need a “glittery” surface to test this. A quite uniform UV-gray surface but with enough unevenness to produce some color noise to compare with the flower. Maybe concrete can do it?

Link to comment
Andy Perrin

Noise is highly temperature dependent, so it's going to be hard to reproduce this. The amount of time the camera had been on, and the ambient temperature, whether it was in the sun or not, all that stuff will affect the internal temp and therefore how much noise you get. It's not as simple as just reproducing the camera settings on a neutral subject.

 

In astrophotography there is a constant fight with camera noise because you are trying to pull a VERY faint signal out of the sky. In that field, they average together many many photos and also take "dark frames" to subtract the effect of fixed noise patterns. You have to take the dark frames at the same time as the original images because otherwise your camera temperature changes and so does the noise pattern. I used the same procedure awhile back on a Queen Anne's Lace, so this is where I learned about it.

Link to comment

If you look closely, there are a ton of little colored dots. To me, that looks like noise.

 

But the back side of this flower did not exhibit any little colored dots even though it did have the weird discolorations shown above. So I don't see how the front side of the flower could show tons of color dots but not the back side given that the photos were made at (almost) the same time with the same gear and the same exposures.

 

Isn't color noise typically more blotchy looking?

Link to comment
Andy Perrin

Ok, look, you have to suggest some kind of mechanism for how this can happen. Current possibilities are:

1) color noise (which I still think is most likely because it's expected to occur, yes even with nice cameras, and even at just ISO200, under the right conditions, especially at high camera temperatures)

2) there is so much IR that it's punching through the Baader somehow.

3) ??

Link to comment

Oh, not to worry, I pretty much agree that it is color noise of some kind. :grin: :grin: I haven't gone off the deep end just yet. :grin:

 

But I did want to post it for discussion because it is kind of odd. It is not the kind of color noise I used to get in UV shots made with that old D200, one of my first conversions. ((I have a lot of D200 botanicals which I never posted because they are so awfully noisy. I should try again with some of those because now we have better converters and de-noisers.))

 

Still thinking about odd demosaicing artifacts due to moire/interference as perhaps possible??, but I'm not glued to that theory.

The back of that Desert Marigold seems to have a fairly complex structure. Maybe I will get a chance to photograph it again this summer and I will look more closely at the back. It is a bit of an odd flower in UV. Here's the formal botanical post. You can see that the first UV Des Mari looks a bit "dusty" in UV. Lots of "dots". Some characteristic of this flower?

Desert Marigold Link.

 

Also thinking about conversion in Photo Ninja where I am almost a knee-jerk applier of the Detail slider. Usually the Detail slider enhances detail (heh!) but sometimes it overemphasized artifacty stuff. I think that has happened for the back-of-the-Marigold photo.

 

Thank you for discussing this with me. It is helpful, for sure!!

Link to comment

My guess: "Mathematical" noise from calculation with small numbers. If you look at the original RGB-values, two are rather large compared to the third one. If you do a WB, you some how multiply each value with a factor, so after that all should have about the same value. As one is small, you need a large factor for this one. So any small error with the small value will get a much bigger error after the WB calculations. Further, there is an influence of whether you do 8bit, 10bit or 16bit calculations.

 

If you look at the "hairs" in the first picture, the color noise after WB increases in the areas, which are darker.

Link to comment
Andy Perrin
That is not a bad theory. I’m sure Andrea uses 16bit workflow but still numerical errors will occur.
Link to comment

That is not a bad theory. I’m sure Andrea uses 16bit workflow but still numerical errors will occur.

Even if a 16-bit workflow is chosen it is difficult to tell how the program's internal calculations are performed.

There are several different ways to do calculations even on 16-bit data.

 

As many calculations of images involve a lot of data, sometimes shortcuts might be taken to speed up processing.

Sometimes our images demand calculations very with rather extreme adjustments.

Even if a calculation-process is OK, with good margins, for normal images, it might not be enough for our odd images.

Link to comment
The 'details slider' in PN is dangerous if you abuse it. It will eventually create a lot of detail noise where nothing was present from the onset.
Link to comment

I once ran a tiff about 15 times through Photo Ninja's detail slider at a high setting.

You get extremely weird results at the end. Lots of fun to try !!

Link to comment

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...