View Full Version : Rolling interference bands
Hello,
I've been testing the video capture function on my camera and ran into some kind of interference issues. Annoying rolling bands appear in the image especially when recording under fluorescent lighting or when there is a monitor in view.
Does anyone know of a way these could be eliminated, or at least reduced? I've looked into DeFreq but I'm not sure if it's meant for this kind of stuff, at least I haven't managed to figure out how to go about it.
Sample clip (http://ajk.pp.fi/video/typing.mkv) encoded in x264, I can provide a shorter clip of the original MJPEG aswell, if it makes any difference.
Thanks!
wonkey_monkey
6th April 2010, 18:08
What shutter speed are you using? Try slowing it down, if you can. I wonder if this is a symptom of a camera with a rolling shutter?
David
I did try a few different shutter speeds but nothing eliminated the bands. The shutter is indeed of the rolling kind, but it is on my camcorder as well, and that doesn't have this banding issue with fluorescent lights or monitors.
Although I do have a separate camcorder, recording video on the DSLR has some benefits with certain kinds of shots due to the variety of lenses available, so I'd like to figure out some remedy for this effect.
2Bdecided
7th April 2010, 12:09
Your camcorder and digital-still-camera-with-video-mode probably have different rolling shutter parameters, even if you adjust all other things to be equal.
Very important is the frame rate.
Most important is the percentage of the total inter-frame time taken to read rolling shutter. The latter isn't a published spec (AFAIK), but it's crucial in determining how much of a problem rolling shutter is. Mobile phones have nearly 100%, CCD is equivalent to almost 0%.
See "Camcorder has a rolling shutter" section, listing values for various cameras, here:
http://www.guthspot.se/video/deshaker.htm
Cheers,
David.
Yeah, I have used Deshaker in the past with my camcorder, and it sorts out the geometric distortion caused by the rolling shutter, i.e. slanted vertical features during pans, quite nicely.
But this is not what is the matter in my sample clip, as the camera is held stationary. The problems are the vertically scrolling bands of varying brightness, which are likely caused by the refresh rate of the monitor (60 Hz) and the lamp (50 Hz). This is what I would need to combat, and my first idea was DeFreq, as this is a type of interference. But I have no idea how to proceed or if there are better tools for tackling the issue.
As far as I can understand, altering the shutter speed will not help here, as the sensor will still be scanned at the same rate, just each line will be given more or less exposure time. The frame rate is 24, which does not match either the monitor or lighting, so...
2Bdecided
7th April 2010, 15:35
Maybe I wasn't clear.
I wasn't proposing deshaker as a solution. I wasn't suggesting deshaker at all. I was pointing to the list of rolling shutter percentages on the deshaker page, which show why two cameras with (supposedly) the same problematic sensor technology can produce different results (i.e. one has the problem, the other doesn't).
I fear there's little you can do about this. Don't shoot under fluorescent lights with that camera. Don't shoot CRT PC monitors with that camera. :(
Shooting 25p, rather than 24p, would at least solve the 50Hz issue. You could maybe set your monitor to 75Hz, and that would work with 25p too.
For this specific footage, mask + temporal soften would work fine - but that's hardly a general fix!
Cheers,
David.
Aw, but where's the fun in giving up straight away... :)
I think I can see the 50 Hz buzz in DeFreq's spectrum view, but haven't managed to do much about it. In principle some kind of notch filter might be able to attenuate the offending bands. Perhaps there's just not enough precision in the signal to work with.
Another idea I had would be to get a clean copy of just the interference, and suitably overlay an animation of it onto the video, but due to the compression artifacts and blocking this would probably be an absolute nightmare to get working properly.
vBulletin® v3.8.11, Copyright ©2000-2025, vBulletin Solutions Inc.