GH2 Skew/Rolling Shutter

HotConductor

HotConductor

Well-known member
Quick question... Rolling shutter occurs because we don't have a universal sensor right? If the problem is simply that the sensor is taking an image from top to bottom, instead of instantly taking snapshots all the time like CCD, then why can't this be correctly easily in post? As in, why hasn't Panasonic or another company developed an algorithm to deal with the vertical refresh? Seems to me that it could be compensated for within software and we'd never have to deal with the jello skew ever again. What am I missing?

Please don't rail on me for asking this question. I think it is legitimate and I am not baiting anyone. I am GENUINELY curious why this problem can't be solved or if there is a solution I simply haven't seen!

This forum rocks. Love my hacked GH2!
 
Even in Vegas I am able to stabilize and 'account for rolling shutter' but it is never a 100% accurate process. It seems to me if we know the math and the amount of time it takes to go from the top to the bottom of the sensor, that there should be a software solution in-camera that would account for this.
 
Hi,
Even though it is not called universal sensor, you ar basically right. The images is progressively recorded from top to bottom.
There are some programs that are able to deal with this issue in some way, but it is just like the smooth camera plugins, it only works in some specific conditions and overall decreases the image quality. There is no way as of right now to fully counter the effect through a program, also some cameras are able to deal with the issue better than other. What the constructor are aiming for is a faster refreshing of the sensor to approach as close as possible to zero.

You can look into After effects or Final Cut Pro X for the correction and see for yourself.
As for now the best way to correct is is to know your limits and work around it. (slower pans, or put a subject to follow in your pan that will prevent the whole image to go sideways)

Good luck.
V.
 
Can't possibly be done properly. You can fix simple manifestations, like -- well, like if you point the camera out the window of a speeding car and shoot a picket fence, then yes, the lean can be corrected -- iMovie even has rolling shutter correction for simple stuff like that.

But what happens if you're shooting that same scene and a truck drives by -- going the other way? Now you have TWO leaning objects, in different directions! How are you supposed to automatically compensate that out?

Furthermore, leaning is only one manifestation of a rolling shutter. Tilting up or down causes the image to stretch out or shrink down, and I don't think you're going to find it too easy to compensate for that. Especially if the camera reverses direction during a single scan, which is entirely possible; you might have the top half of the frame squished down and the bottom half stretched out.

The solution for rolling shutter is to go to a global shutter as soon as practical and affordable, and, barring that, to just scan the chip as fast as possible. On the more-expensive cameras, that's what they do: the AF100/FS100/Red One/EX1/C300 etc all have much much less rolling shutter than the GH2 (when both are in 24p mode).

Interestingly on the GH2, you can see a huge difference in rolling shutter performance between its 1080/60i or 720/60p settings, and its 1080/24 mode. 1080/24 has the most severe rolling shutter effect, and the 720/60 or 1080/60 are actually pretty good.

Finally, if you want to minimize rolling shutter on a GH2, try shooting in ETC mode, the ETC scans the chip much faster and therefore the rolling shutter effect is greatly minimized.
 
Basically, the issue is the relative motion of things in front of the sensor are not the same. If you are shooting a static shot of say a building, the building is not moving. But, if a bus goes by, you'll see slant in the bus because it is moving, yet, the building is not moving so it remains straight. There are some algorithms I've seen that actually do try and account for this, but really, they are VERY complicated and hugely prone to artifacts. Trying to globally fix the bus going by makes the building slant.

If you are doing a whip pan and everything is stationary, you could fix it in post decent enough. Of course, you'll be loosing resolution because you'll need to blow up the picture after you chop off the sides since you've now create a parallelogram :)
 
Barry_Green said:
Finally, if you want to minimize rolling shutter on a GH2, try shooting in ETC mode, the ETC scans the chip much faster and therefore the rolling shutter effect is greatly minimized.
Click to expand...
It would be interesting to see if with all these new hacks someone could implement the faster ETC scan rate into the normal standard sized footage.
Then again it's probably not scanning "faster" but just having to scan less than full sensor mode.
 
Could we duct-tape two GH2s (one up, one down), take the image, and... just joking.

I appreciate the elaboration and I understand the variations. My only point is that since the math is consistent, it seems like we would be able to compensate for it. BUT, like you all said, the movements are not constant on all the objects, so it isn't so simple. 4 CMOS sensors (0, 90, 180, 270 degree marks), combine them all, use math = universal sensor!!
 
4 CMOS sensors (0, 90, 180, 270 degree marks), combine them all, use math = universal sensor!!
Click to expand...
4x sensor = 4x PRICE : ) kinda like the price diff between a 7D and 5Dll, only add it three more times.

i hear rumor that panasonic MAY be working on a global shutter for the GH3... that might change my tune on my whole hpx170 and my cold dead hands thing!
 
Last edited:
You must log in or register to reply here.
Back
Top