Sensor pattern noise issue.

Has anyone ever encountered noise like this?

56mm 1/640 f1.8 ISO200

Crop of an image taken on my Olympus Em1-II with a Sigma 56mm f1.4. Seems to happen when shooting in pretty specific backlit situations. I can recreate it with my 40-150 f2.8 as well. Haven't been able to get it with my 12-40 f2.8. Happens with and without a UV filter.

I'm guessing it's similar to red dot flare, light bouncing off the sensor and rear element, but it would be just a guess.

I can avoid it with technique, but sometimes it's a nice angle for shooting portraits. Just curious if anyone has ever seen this before and knows what the cause is.

Is this a SOOC JPEG, or was this a converted Raw file. If converted raw it may have been introduced during the raw conversion.

Gao Gao wrote:

MFotoAus wrote:

It had crossed my mind, I think it's too dense for the pdaf though. Don't know.

There was this thread on PDAF artifacts:

https://www.dpreview.com/forums/thread/4108353

Looks like it.

I think you nailed it. Bill Claff later examined another picture with an identical pattern and said it looked like a PDAF pattern after some examination (initially he thought it was something else also). Given the same camera and what looks like the exact same pattern, I would say it's likely PDAF also.

https://www.dpreview.com/forums/post/60209440

The conditions for it showing up also matches when PDAF grids show up in many cameras (namely backlit conditions).

Yes, it definitively looks like PDAF sensors.

So there might actually be a good reason why Panasonic develops its own approach to autofocus 😀 (wink to the naysayers).

I have photographed extensively - even regularly! - back lit scenes and subjects, because I like the resulting contrast and silhouettes. I have NEVER seen anything even remotely close to this effect, though my images are a much, much strong contrast situation.

This camera is now about 4 years old, and I never read anything about this effect.

To me, it is so bad that it would lead me to send the camera for repair or even abandon a camera system if i were to find it regularly. It is a horrible picture!

Is the camera fine?

MFotoAus wrote:

This was a RAW conversion in Lightroom, but the effect is so strong it can be seen in camera as well.

Thanks, I have seen pattern noise routinely when .Orf files are converted directly in Lightroom, but that noised has a signature look to it that is much different than what you are showing. I have been shooting PDAF with Olympus since the original  E-M1days and have never seen what you are presenting......a curiosity, for sure.

Gao Gao wrote:

MFotoAus wrote:

Ahh, thank you. I see now. Never realised there was that much detail in the array. Always learning. Now I just have to work on avoiding it.

I'm still curious about the exact mechanism of such artifacts, and whether it's possible to fix it in software after the fact.

Do you think you can share a sample raw photo?

Sony's have a similar issue that shows up as striping and a tool was developed to do repairs. RawTherapee also has a fix built in for it.

http://aggregate.org/DIT/KARWY-SR/

http://www.aggregate.org/DIT/EI2019/karwysr.pdf

General idea is identifying which pixels have the issue and then overwriting them with interpolated pixels.

However, the pattern is different here from the stripes so that tool might not work here.

The closest situation I see to this is watermark removal. I would guess an approach would be in Photoshop to build a pattern layer (somehow enhance the pattern so you have a layer that is only that pattern) then subtract that from the image. That of course requires the effect to be uniform across the image, which it might not be.

Gao Gao wrote:

JakeJY wrote:

Gao Gao wrote:

MFotoAus wrote:

Ahh, thank you. I see now. Never realised there was that much detail in the array. Always learning. Now I just have to work on avoiding it.

I'm still curious about the exact mechanism of such artifacts, and whether it's possible to fix it in software after the fact.

Do you think you can share a sample raw photo?

Sony's have a similar issue that shows up as striping and a tool was developed to do repairs. RawTherapee also has a fix built in for it.

http://aggregate.org/DIT/KARWY-SR/

http://www.aggregate.org/DIT/EI2019/karwysr.pdf

General idea is identifying which pixels have the issue and then overwriting them with interpolated pixels.

However, the pattern is different here from the stripes so that tool might not work here.

The closest situation I see to this is watermark removal. I would guess an approach would be in Photoshop to build a pattern layer (somehow enhance the pattern so you have a layer that is only that pattern) then subtract that from the image. That of course requires the effect to be uniform across the image, which it might not be.

Thanks for the links. I haven't seen a raw file yet but I suppose fixing it before demosaic should be better than after. If the fix is only for E-M1 Marks II and III, I could see using DCRAW as the basis and generating a repaired raw for any raw developing software.

Definitely, if possible to do before demosaic, the better (that's how that tool works). Since you are interpolating pixels anyways to demosaic, doing it before then gives you even better chances of hiding the effect. But working with the RAW that way is going to be outside of the realm of most photographers, and most would be more familiar with tools that work after demosaic.