Dmitry Shijan wrote:Here is my summary:
1. Wrong debayering of second green pixel produces semy-invisible fixed pixel grid which is partially hidden under the sensor noise. As a result at some zoom or scale levels in various players and apps this invisible grid may produce highly visible cross-hatch moire-like pattern.
2. Depending of image contrast, added sharpen and noise reduction amount, and player scaling method this pattern may be visible or not visible at all.
3. In BMMCC which is limited to 1920x1080 resolution this makes sensor Noise and FPN more visible (it behaves like some kind of shadow noise and FPN sharpening, adds additional digital artifacts to noise texture) and this limits the amount of visually usable shadow details.
4. People noticed this effect in BMPCC and BMPC cameras, but there is no clear tests yet. In BMPCC it is harder to see because it is slightly noisier than BMMCC and pixel pattern more hidden under the noise.
5. This effect depends of camera firmware version.
6. This problem can be fixed with fix in CornerFix app, but this fix works only in ACR
https://sites.google.com/site/cornerfix ... patterns-17. BM support team answer: "this is currently under investigation by our Development team"
Hey Dimitry,
Thanks for your summary, I think you and Earl are hitting the nail on the head to be honest, I have a limited understanding of debayering and I'm actively trying to learn about it, I have been grabbing lots of footage together to do a little screen cast on what I have found on my UM 4.6k.
I wasn't too fussed about the scaling crosshatching issue as I knew that wasn't always visible and it didn't transfer over to an exported H.264 (probably hidden with the loss of detail in the compression), however after submitting my report to BM support and filming that gig on the same evening when I got back I was far from impressed with the footage at 4.6k.
It was tough lighting conditions for sure however something just isn't quite right with the way the sensor is handling the images at 4k resolutions IMO, you get away with it in good lighting situations and to be honest examining my shots in good light the pattern in the shadows is negligible really, you can only see it when really blowing up the image which is just stupid as people will never view it like that.
An interesting test I did today was to put a well lit shot together with a poorly lit scene from the gig, played at full screen the well lit shot shows no cross hatching and as soon as the shot cuts to the dark scene bang, cross hatch hell, so as Earl has mentioned, the scaling is the bi product of the small pattern within the dark areas of the image, people shooting RAW seem to be able to correct this however I never shoot RAW so the in camera debayering needs some sort of fix when encoding to ProRes.
I really sound like I am moaning all the time on here and I really don't want to, I waited a long time for this camera and in good light 4k is defiantly usable however in more challenging situations I would probably now opt for no higher than 2k.
It's great to hear that its under investigation and it would be great to hear an official response about them working on this, overall I have been very happy with the images I have been getting in good light and although its a big investment for me personally I have to remember this camera body is £3,500, so its going to have its pitfalls.