5

A few days ago I took a night sky image of the milky way with my XPERIA 1 II and wanted to take a closer look at the RAW file in Lightroom Mobile. When doing so, I realised a strange circular pattern. Though "auto" correction already revealed this pattern, I additionally increased the exposure to make it even more visible.

night sky image with circular pattern
ƒ/2,2 30 Sek. 2,67 mm ISO 50

For me, it looks like something from the lens, but I am very new to photography and especially in post-processing RAW files, so I am wondering what it might be.

Is this something that can be corrected by whatever additional images?

13
  • 1
    You say autocorrection revealed the pattern. What does it look like without autocorrection?
    – scottbb
    Sep 1 '20 at 14:40
  • @scottbb Nearly black, with some stars in white and a quite green tree. Sep 1 '20 at 15:02
  • @RomeoNinov Not sure, but at least it's a good guide. From a short calculation, I'm quite sure the RAW file has 16bit colour depth. Sep 1 '20 at 15:02
  • @pH13-YetanotherPhilipp, I have my doubts about 16 bits. Most DSLRs have 14 bits, few hightend cameras (which cost 20k or more) have 16 bits. You will be happy if you have real 13 bits. And this is normal situation (when you pump up shadows). How do you expect to make acceptable night images with such small sensor? Sep 1 '20 at 15:19
2

I do not think this artifact is caused by reflections in the optical path. I think this is due to software correction of light falloff as a function of distance from image center (i.e., vignetting compensation). Normally such correction is not very noticeable because the signal-to-noise ratio (SNR) is high enough. But in this case, the image is so dark, that lifting the the image by a couple stops towards the corners greatly magnifies the noise in those regions as well.

Notice that the noise at the top, above the outer ring, and in the bottom center and lower right, is substantially more pronounced ("grainier") than in the center. There is nothing optically that would cause radially-biased digital noise. That has to be a result of software increasing the signal level (i.e., digital multiplication). The upper-right and upper-left corners outside of the ring especially show a lot more noise. Optically, these regions should be darkest due to vignetting, so their SNR is lowest. After digitally boosting the corners for light falloff, the only signal there is just boosted noise.

5
  • That would have to be the worst vignette correction algorithm ever created... and it doesn't account for the corner being overlayed with another artifact. Sep 16 '20 at 13:58
  • @StevenKersting Not really. Don't forget, the original image was probably really dark. So the max signal values in the darkest region were in the lowest few bits. And the noise amplitude in those bits was probably 50%, perhaps even 1:1 SNR in some dark places. So there's not much an algorithm can do but boost noise. I'm sure the algorithm is trying to smoothly increase the brightness radially, but we're seeing the quantization stair step because the signal is so low. i.e., it might be trying to multiply the value by 1.25, but if the signal is only 7, there's going to be truncation, which is...
    – scottbb
    Sep 16 '20 at 22:07
  • 1
    ... a significant error from the actual desired value after adjustment. Re: the corner: I just assumed it's overhang over the house, and the bright spot along the corner border is a low-level reflection off the rain gutter. That's what it looks like to me.
    – scottbb
    Sep 16 '20 at 22:09
  • I have never seen a vignette correction generate rings... they generate a gradient. I could accept banding as part of the gradient/correction; but not with the distinct/dissimilar bands/rings. Sep 17 '20 at 2:09
  • 1
    @StevenKersting it would only generate rings if the signal level is very low. You can't create a gradient from only a few bits of data. If the signal value is only, say, 0-7 (3 bits of data), you can't generate something like 6.25. So it gets truncated and posterized. Don't forget that OP's image had sliders cranked way up in post. The original image was very dark. If the picture were in daytime, you wouldn't see rings like that, because the sky value would probably be something like 50-75% full scale, which would take a gradient nicely.
    – scottbb
    Sep 17 '20 at 5:01
6

That very much looks like a reflection of the lens on the protective glass to me... along with another reflection from something outside. It does not look like banding due to editing.

This is the kind of thing one might see with an SLR lens that has an uncoated filter on it.

IDT there is anything you can really do about it. In more normal conditions it is not likely to be problematic as there is much more light; which makes the dark reflection on the glass inconsequential in the total exposure.

enter image description here

It matches up a lot like this enter image description here

3
  • 1
    That's a smart idea! That explains the circular pattern very well. Sep 1 '20 at 18:23
  • But if it was made by reflections, shouldn't there be any irregularities depending on the position of the main light source, which in this case would probably be the tree reflecting the light from inside? Sep 1 '20 at 19:14
  • 1
    @pH13-YetanotherPhilipp, what do you mean by irregularities? Something like a catchlight showing the source direction? I think the reflection of mostly/nearly black lens/surround just didn't contribute enough to be that apparent. That's why it's only apparent at all with recovery. Sep 1 '20 at 21:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.