Strange noise on the background, appears after processing raw images

[Hoka Hey]

I have old lightroom 6. Can I reproduce this with any of my noisy images and a lens profile?

I don’t know what you can do with LR 6. I use the current versions of ACR and PS.

Is it a specific lens profile?

This becomes visible when you try and push underexposed parts of a file too far in ACR or LR with the lens profile turned on.

I haven’t tested every lens, but I don’t think that it is lens dependent.

Well, I guess it's lens dependent to the extent that the lens profiles are lens dependent. So, I would expect it to be worse with lenses that need more correction.

Point taken.

Just thinking back without spending a ton of time going back and finding the files, I think that it shows up in most files when stretched past their limit. The patterns just differ.

 

[CatchingTime]

It is not moire, that pattern is from the lens profile that was applied. De-select it, and you'll see the pattern go away.

I don't think that the lens profile correction is critical for images like this, as any peripheral shading correction can be applied using the generic slider, and it is not an architectural-type image where distortion needs to be corrected.

I assume the profile in Lightroom. What is it interfering with to cause the 'fringing'. The original pixels I assume were totally replaced by the remapping of the correction.

Although it looks like an interference pattern, I honestly don't know. I wouldn't think that the original pixels were (actually) replaced, but rather there was some sort of stretching/compressing of brightness values that somehow results in effectively lowering the light in the 'rings', thus increasing the noise.

I've known about this issue for years and it's mostly easily avoided by not applying lens corrections when they really don't do anything constructive (as in the example shown above). I honestly never looked it up to find the technical reason it occurs, but I'm sure someone has delved into it...

 

[Mike Engles]

The star images are consistent with distortion correction.

The original has considerable horizontal noise banding.

The distortion corrected one has considerable circular noise banding. The correction has stretched the image from the centre outwards progressively more towards the edges. The correction also uses nearest neighbour interpolation, so makes the displaced noise even more pronounced.Try using the spherize in Photoshop, That is what the distortion correction is essentially doing.

The concert picture has no noticeable horizontal noise. It has a very complex pattern, much more consistent with moire. For the correction to be responsible for the pattern, should also mean that the image will have been totally mashed up and it is not. It does not say very much about the lens if it needs such an extreme correction

As far as I know we have not seen a similar image to the distortion corrected one, but without the distortion correction, using Lightroom.

 

[Bigger]

It is not moire, that pattern is from the lens profile that was applied. De-select it, and you'll see the pattern go away.

I don't think that the lens profile correction is critical for images like this, as any peripheral shading correction can be applied using the generic slider, and it is not an architectural-type image where distortion needs to be corrected.

I assume the profile in Lightroom. What is it interfering with to cause the 'fringing'. The original pixels I assume were totally replaced by the remapping of the correction.

Although it looks like an interference pattern, I honestly don't know. I wouldn't think that the original pixels were (actually) replaced, but rather there was some sort of stretching/compressing of brightness values that somehow results in effectively lowering the light in the 'rings', thus increasing the noise.

I've known about this issue for years and it's mostly easily avoided by not applying lens corrections when they really don't do anything constructive (as in the example shown above). I honestly never looked it up to find the technical reason it occurs, but I'm sure someone has delved into it...

Yes, but isn't it possible that the lens correction is just amplifying artifacts from the debayering? These are normally invisible, but if you push the shadows too far, they become visible.

We really need the RAW file.

 

[Mike Engles]

It is not moire, that pattern is from the lens profile that was applied. De-select it, and you'll see the pattern go away.

I don't think that the lens profile correction is critical for images like this, as any peripheral shading correction can be applied using the generic slider, and it is not an architectural-type image where distortion needs to be corrected.

I assume the profile in Lightroom. What is it interfering with to cause the 'fringing'. The original pixels I assume were totally replaced by the remapping of the correction.

Although it looks like an interference pattern, I honestly don't know. I wouldn't think that the original pixels were (actually) replaced, but rather there was some sort of stretching/compressing of brightness values that somehow results in effectively lowering the light in the 'rings', thus increasing the noise.

I've known about this issue for years and it's mostly easily avoided by not applying lens corrections when they really don't do anything constructive (as in the example shown above). I honestly never looked it up to find the technical reason it occurs, but I'm sure someone has delved into it...

Yes, but isn't it possible that the lens correction is just amplifying artifacts from the debayering? These are normally invisible, but if you push the shadows too far, they become visible.

We really need the RAW file.

Try as I might I am having great difficulty in actually reproducing the effects in the sky picture or the concert one.

As for distortion, yes since correction remaps the pixels (locally and variably enlarged) and then since the image now has holes in it, it uses nearest neighbour to interpolate, so artefacts can become magnified.

 

[Hoka Hey]

It is not moire, that pattern is from the lens profile that was applied. De-select it, and you'll see the pattern go away.

I don't think that the lens profile correction is critical for images like this, as any peripheral shading correction can be applied using the generic slider, and it is not an architectural-type image where distortion needs to be corrected.

I assume the profile in Lightroom. What is it interfering with to cause the 'fringing'. The original pixels I assume were totally replaced by the remapping of the correction.

Although it looks like an interference pattern, I honestly don't know. I wouldn't think that the original pixels were (actually) replaced, but rather there was some sort of stretching/compressing of brightness values that somehow results in effectively lowering the light in the 'rings', thus increasing the noise.

I've known about this issue for years and it's mostly easily avoided by not applying lens corrections when they really don't do anything constructive (as in the example shown above). I honestly never looked it up to find the technical reason it occurs, but I'm sure someone has delved into it...

Yes, but isn't it possible that the lens correction is just amplifying artifacts from the debayering? These are normally invisible, but if you push the shadows too far, they become visible.

We really need the RAW file.

Try as I might I am having great difficulty in actually reproducing the effects in the sky picture or the concert one.

How are you getting the images into LR 6? Does it support the R5?

As for distortion, yes since correction remaps the pixels (locally and variably enlarged) and then since the image now has holes in it, it uses nearest neighbour to interpolate, so artefacts can become magnifie

 

[Mike Engles]

No I am trying with my own images using my 100-400 images and high ISO

I have looked at about 20 and all I see is that with correction the pixels are shifted outwards. No smearing or extra patterns The concert picture is at 260mm, I don't have any at that focal length.

That pattern is so distinct that it has to be moire. albeit at very low level as any lights facing the sensor are low. The RAW in Lightroom with all corrections, but distortion will be definitive

 

[Hoka Hey]

No I am trying with my own images using my 100-400 images and high ISO

Sorry. I don't understand your reply.

If you are debayering the files before they get to LR, then I don't think that you will be able to reproduce the effect.

I have looked at about 20 and all I see is that with correction the pixels are shifted outwards. No smearing or extra patterns The concert picture is at 260mm, I don't have any at that focal length.

That pattern is so distinct that it has to be moire. albeit at very low level as any lights facing the sensor are low. The RAW in Lightroom with all corrections, but distortion will be definitive

I already provided you with files above that were with and without distortion correction. These were the first set of images that I posted of the comet. You only see the effect in the ones with the distortion correction turned on.

 

[Mike Engles]

I have already addressed those files, earlier. I am trying using my own files with the corrections for the 100-400 in Lightroom, to see if I can get a result anything like those. I cannot.

The Comet files could be due to distortion correction, the concert file has nothing to do with distortion correction. It is moire.

 

[John Sheehy]

So these are digital artefacts caused two by 2 over laid grids interfering, because that is as far as I know, is the only way to cause moire.

"Moiré" or not by an definition, this is a problem of quick-and-dirty resampling combined with periodic luminance "noise". There are likely two different blackpoints for the two interleaved RAW green channels, and the lens correction does not weight them well when stretching the image.

When you have uneven green channels, and you do not stretch the image, the pattern is so fine that it disappears in your vision, just like a simple repeating dither. The converter may even suppress this when it is fine. When you don't weight a stretch properly, then you have slow cycles, several pixels or more, where the influence goes from mostly one green channel, to the other, and back again, and this appears as large-scale detail, which is not suppressed by either software or vision.

 

[John Sheehy]

Does not distortion correction replace one set of pixels with another. All pixels are remapped by a transformation. Pixel A (xy) is transformed to A (x1 y1).This is an interference pattern, which need two grids. Where are they?

As above, the Bayer pattern and the lens profile map.

It could happen with a monochrome sensor, too, if the blackpoints were different in a repeating periodic pattern, and especially if quick-and-dirty resampling was used.

 

[Mike Engles]

So you are saying that this is not a interaction to the RAW file of 2 grids interfering at acquisition, that is the sensor grid and a lighting array of LEDS, classic moire,

You are saying that it is an interaction by Lightroom decoding the RAW file, a grid and the lens correction profile, another grid. Also classic moire

I really do not know, but I have been trying to reproduce this 'software' moire, using my old Canon RAW files in Lightroom using the Lightroom correction profile for my lens (the same as the concert) and cannot make that pattern or anything like it. The pixels just shift from the centre to the edges, as they should. If it is what you say, it should be reproducible, unless it is a property of R5 files only. So I tried the same with R5 files and also cannot see any pattern. I use Lightroom 6 with old lens profiles.

I contend that this is the first case, but the pattern is really ill defined, as the lighting array was dimmed. Bringing up the shadows makes the pattern more pronounced and then the distortion correction does its work, shifting the pixels and filling in the holes, with nearest neighbour interpolation, making the pattern even more pronounced.

I think either way way it is classic moire.

The comet image is different again. That image has been partly rotated (long shutter speed with a 70mm lens) and then correction added. That is why the noise is.'swirled' as it is a bit before correction. The interpolation makes it more pronounced.

 

[Mako2011]

I think either way way it is classic moire.

That's really not the case here, IMO. John Sheehy is usually dead on. Here as well I think

 

[Mike Engles]

If he is correct, then the effect should be able to be reproduced consistently.

Hypotheses are only valid if confirmed by experiment

 

[Mako2011]

If he is correct, then the effect should be able to be reproduced consistently.

Hypotheses are only valid if confirmed by experiment

Never the less...pretty obvious it's not classic moire

 

[CatchingTime]

It's amazing how a thread can devolve into a multi-page debate when the answer was given on Page 1. It is not classic moire, but it is an example of an interference pattern caused by the LR lens profile being applied and is 100% reproducible, the effects of which are enhanced by high ISO images when opening up the shadows:

• If your image doesn't have large areas of dark tones, it will be harder to see;
• If your image has lots of detail, it will be hard to see;
• If your image is shot at lower ISOs, the effect is harder to see;
• If you have not opened up the shadows, the effect is hard to see;
• If your camera has good high-ISO noise performance, the effect is harder to see;
• If you use a lens with very little distortion, it will be harder to see.

It's essentially posterization due to stretching very low signal in the shadows, and is most apparent in the corners due to the correction of peripheral shading. The 'shadows' in question need to be nearly black to highlight the issue. If you stack images (blend:lighten), the effect becomes more pronounced. Ask DSO astro guys if they apply lens profiles prior to stacking.

Different lenses obviously (maybe not so obvious?) produce different patterns.

https://starcircleacademy.com/2013/12/the-revenge-of-lens-correction/


Canon R5, real-world image; better noise performance, higher exposure, thus rings are harder to see.


Canon 5Ds + Sigma 14/1.8, with and without lens profile applied (first one is WITHOUT lens profile).


Canon 5Ds + EF16-35/4L

--
My $0.02 in a world where pennies are obsolete.

 

[Hoka Hey]

If he is correct, then the effect should be able to be reproduced consistently.

Hypotheses are only valid if confirmed by experiment

I can reproduce it consistently with images from the DPR sample gallery for different cameras from different brands. I don't know why you can't reproduce it, but I'll offer once again to test your images if you will provide a link to some of your RAW files that are underexposed and have an area that is in deep deep shadow.

 

[CatchingTime]

If he is correct, then the effect should be able to be reproduced consistently.

Hypotheses are only valid if confirmed by experiment

Anyone who isn't convinced, take a long exposure at high-ISO, wide open with the lens cap on. Open the image up by +5 stops, and then compare with/without lens profile applied.

For even more fun, do this with all your lenses (especially wides), and marvel at the different patterns due to each lens's specific distortion and peripheral shading characteristics (for even more fun, add some psilocybin into the mix!).

If you then agree, you can go back to my first post and give it a thumbs-up.

If you don't, you can continue arguing, but I'll likely move on to the next thing on my list...

--
My $0.02 in a world where pennies are obsolete.

 

[Mike Engles]

Well I have a picture of the moon 5D2 RAW, a lot of black and noise. I pushed the shadows as well as brightness and applied in Lightrroom the Adobe profile for my 100-400 lens at 400 mm. I have panned all over this and just see pixels move as I switch between correction and no correction. I see no interference patterns anywhere.

None at all like the concert picture, nor the circular smearing in the comet picture. What am I missing?

No correction



Correction







tion

If this is the result of decoding digital images and lens profile correction, it eludes me. Take one of your images and go through the steps, like I have.

 

[CatchingTime]

Well I have a picture of the moon 5D2 RAW, a lot of black and noise. I pushed the shadows as well as brightness and applied in Lightrroom the Adobe profile for my 100-400 lens at 400 mm. I have panned all over this and just see pixels move as I switch between correction and no correction. I see no interference patterns anywhere.

None at all like the concert picture, nor the circular smearing in the comet picture. What am I missing?

No correction

If this is the result of decoding digital images and lens profile correction, it eludes me. Take one of your images and go through the steps, like I have.

You're missing a lens with ample distortion shot wide open, and possibly a camera with poor high-ISO noise characteristics. I'm sure you know that telephoto lenses have much, much less distortion than UWA lenses, and any lens shot 2+ stops from wide open will have less vignetting. Pick a wide lens, shoot it wide open (with lens cap on), and repeat. Also, viewing at 33% in LR vs. 25% (etc.) can show the interference pattern, which gets hidden with downsampling (by the graphics card).

You can do it right now, sitting at your desk. Just follow my instructions above.

--
My $0.02 in a world where pennies are obsolete.