The Math Behind the Rolling Shutter Effect.
October 13, 2014 7:33 AM   Subscribe

Here's a pair of blog posts explaining the math behind the "Rolling Shutter Effect": Playing Detective with Rolling Shutter Photos and Rolling Shutters.
posted by empath (12 comments total) 26 users marked this as a favorite
 
Some background: Rolling Shutter (wikipedia). This effect is something that camera folks have been battling and obsessing over for years now. With the rise of DSLRs, every filmmaker, pro and amateur is familiar with these effects. Newer cameras are going to incorporate global shutter, so rolling shutter will eventually be something confined to earlier tech. Here's an article about the two all the way back in 2009 (an eternity in cam tech): The Global Shutter / Rolling Shutter Debate.
posted by VikingSword at 9:27 AM on October 13, 2014


Previously, without math and spiffy GIFs: The focal plane shutter artifact in certain digital cameras.
posted by filthy light thief at 9:36 AM on October 13, 2014


Surely DSLRs are among the types of digital camera that don't have rolling shutter effects?
posted by edd at 9:42 AM on October 13, 2014


edd: why would lens configuration change how digital sensors work?
posted by idiopath at 9:54 AM on October 13, 2014 [1 favorite]


DSLRs still have it when shooting video, because they don't use a mechanical shutter at that time. They're pulling data off the sensor just like an iPhone does, so they'll still run into the problem. Not sure if they can avoid problems like this though. The sensor is still the same kind of sensor, whether it's a big one or a small one, or if there's a mechanical shutter in there or not.
posted by AzraelBrown at 9:56 AM on October 13, 2014


Propellers not required - this is most commonly seen when you pan the camera quickly and everything becomes jello.
posted by echo target at 10:17 AM on October 13, 2014 [1 favorite]


Idiopath: it's not the lens configuration - it's the mechanical shutter that is so fast it can allow a fairly uniform 1/4000s exposure or shorter.

I obviously accept it still applies to a DSLR in video mode.

Somewhat relatedly, older CCD chips read out from just one corner, and you slide the charge off in columns and rows from that corner. The Sloan Digital Sky Survey used to image the sky by pointing at one location relative to the ground, and (with a rotator unit) slide charge off so that the entire exposure slid along the chip at the same rate the sky moved overhead, thereby taking long strips of the sky. Clever method.
posted by edd at 10:43 AM on October 13, 2014




I remember first learning about rolling shutter artifacts from this old web page by a guy who converted his flatbed scanner into a camera, producing this memorable photo of his garage door.

"I obviously accept it still applies to a DSLR in video mode."

As AzraelBrown points out, with still images it could still be a problem for any motion in the image that's faster than the shutter speed, though depending upon the CMOS. I haven't found any discussion of this with googling, however.

"The Sloan Digital Sky Survey used to image the sky by pointing at one location relative to the ground, and (with a rotator unit) slide charge off so that the entire exposure slid along the chip at the same rate the sky moved overhead, thereby taking long strips of the sky. Clever method."

I visited Apache Point Observatory in 1995 when SDSS was being built and they talked at great length about the CCD, which was pretty advanced for the time. Some of the people I was with were affiliated with NRAO/VLA and they were all pretty impressed with SDSS, particularly the computing/networking aspect of it. After the visit, I had an inkling of how important SDSS was likely to be and it's been very cool over the years to see the science just keep coming from it.
posted by Ivan Fyodorovich at 12:49 PM on October 13, 2014 [2 favorites]


Ah, the rolling shutter artifact! It gets a lot more play than its less-bizarre companion, the flash band. I must have been the only guy watching the media-feeding-frenzy scenes in Gone Girl last week and thinking, "Man, wish they had figured out a way to tame those flash-band artifacts" as the camera flashes popped — and lit up only part of the screen.
posted by Mothlight at 2:14 PM on October 13, 2014


As AzraelBrown points out, with still images it could still be a problem for any motion in the image that's faster than the shutter speed, though depending upon the CMOS. I haven't found any discussion of this with googling, however.
Sure, but it's a lot faster than the rolling shutter effects seen in propellers and the like, and aren't then a feature of digital cameras but the mechanical shutters used.
posted by edd at 4:23 PM on October 13, 2014


with still images it could still be a problem for any motion in the image that's faster than the shutter speed

It's a bit different when it's an all mechanical shutter. The camera first empties all the pixels of their charges, then opens the mechanical shutter. The light creates a charge on each pixel. When enough light has come in, the shutter closes and all the pixels stop charging. While the shutter is closed, the camera goes through and reads out the level of each of the pixels. The readout is happening in the same sequential way that produces rolling shutter, but since no new light is being captured during the readout you won't get any artifacts.

Anything that moves during a mechanical exposure just gets blurred in the familiar way. AzraelBrown's photo wouldn't be possible with a camera with a mechanical shutter, which includes DSLRs as well as mirrorless and compact cameras.
posted by echo target at 8:25 PM on October 13, 2014 [2 favorites]


« Older Star Trek Fact Check   |   Crack, the CIA, and the Contras Newer »


This thread has been archived and is closed to new comments