@Manuel_Angel - thanks for supplying these images from your setup! The circle of people having a little bit of noise in their captures widens! Here’s the red channel of your capture with the shadows pushed up into medium gray:
As you can see, there is a tiny bit of noise on the right side of the frame. Very similar to my captures!
Your empty film gate image shows no banding at all - that was what I was suggesting: the noise is only present in the dark areas of the image, not present in medium or high intensity ranges.
@jankaiser - thanks as well for joining the group!
All of what you listed here gets rid of this noise:
“lowered the exposure”, i.e. keeping the dark shadows dark, not pushing them too much into medium brightness levels.
allowing me to rephrase “processed video frames” into “encoding into H264/H265”: video compression usually hides this noise quite effective. It simply gets kicked out by the compression algorithm.
“colour noise reduction step” is my favorite angle of attack. Spatio-temporal denoising gets rid of the noisestripes for sure.
Many thanks to both @Manuel_Angel and @jankaiser for your fast reaction! Taking into account the diversity of the scanners, their difference of the surroundings with respect to EMI and the various illumination sources employed, by noting the similarity of the noise patterns, I think the verdict is clear: this noise is mostly a peculiar property of the IMX477 sensor chip used in the HQ camera.
@cpixip Indeed, I must admit that my camera also generates noise bands. I had never imagined this. In fact, in normal captures they go completely unnoticed.
Thank you for your always interesting contributions.
…
Interestingly, the processed video frame does not show this effect, though it is possible I lowered the exposure slightly to hide weird effects … I don’t remember. It also went through a colour noise reduction step.
This is the first capture that really does look like the banding issue I’m complaining about. Thanks for sharing.
I also found out that the banding can be made less visible (or even invisible) by applying color noise reduction, but I’ve a had a few scenes where noise reduction produced a bunch of weird artifacts and became hesitant to use too much of it…
Well. I think I might finalize one of the really “bandy” scenes and see how bad it ultimately is. I’m glad to see it’s not necessarily my setup at fault. Many thanks for all the contributions .
Which noise reducer are you using? While I love NeatVideo’s temporal noise reduction, I have since learned that enabling the “Dust and Scratches” feature in it generates a ton of artifacts near any fast motion, making the results basically unusable. It took me a while to diagnose the source of the artifacts, but now I just leave that box unchecked and it works flawlessly.
If you’re using something similar, maybe there is some subset of features you can enable to keep the noise reduction but avoid the artifacts?
Which noise reducer are you using? While I love NeatVideo’s temporal noise reduction, I have since learned that enabling the “Dust and Scratches” feature in it generates a ton of artifacts near any fast motion
DaVinci only. But you’re right. The worst artifacts there also come from dust removal if you make the effect too strong.
I took two scenes and treated them with color noise reduction. It was okay-ish in a scene with a road in deep shadows, but the above sunset turns out really bad. The effect needs to be so strong that it desaturates large parts of the frame, which is alright for a shadowy road, but not okay in a glowing sunset.
But I’ve found a dark piece of film from a strip of scene transitions and will try to set up the exposure based on that - as well as checking whether restarting the camera after a while makes any difference. I’ve considered switching to an external camera and using gphoto, but (at least with my old Canon) gphoto seems to be really slow (0.5fps at best) - and then I’d also have to find an adequately fast camera with interchangeable lenses and electronic shutter at the cheap …
From the above findings @cpixip established that everyone using the HQ, with particular conditions, may encounter the dark/red channel banding. The hypothesis is that it is caused by the color processing pipeline, in particular by the underside clipping of noise, and giving that red is the noisier channel, said under clipping creates may be the cause of these. Until we figure out a workaround, it appears inherent to using the HQ.
In the case of @verlakasalt, there is an additional banding from the light. Here is a visualization using imagej of this frame.
I would point out that this light-banding would also be in the film dark areas, and I would speculate that would make the clipping-banding more visible.
Edit
It is also worth pointing that while banding would be equally in the film dark and lid areas, it would be perceived as more intense in the dark areas.
Taking a slice of the dark left area of the beach shot (even when there is some sky)
Fixing the light would not eliminate the issues shown in the shots capped. However, I would speculate that the light is a significant contributor that is making these much worst, particularly in dark film areas which would have the worst black-clipping banding + light-induced banding.
I suspect I should’ve gone with a constant current strip? Can I just use a different power source? @Manuel_Angel How do you power your lamp?
@cpixip: I tried restarting the camera (assuming that camera.stop() is the correct command), but it didn’t change anything about the banding characteristics.
The lamp I use in my device can be powered between 10 and 18 V.
To power my scanner (motor and lighting) I am using a 12V power supply that I salvaged from an old analog flat screen TV from 2005.
… I don’t remember. It also went through a colour noise reduction step.
.
