What's this dancing black frame thing?
June 4, 2012 12:03 AM   Subscribe

Can somebody tell me what's going on in this video? As much as I feel terrible for the people there, I'm not talking about the scene(s) being shown, but rather the dancing black frame around the edges!

Is this a thing? Does it have a name? What is it actually?

At first I thought it was someone filming a TV or monitor without a tripod, but that doesn't seem to be the case.

Please help! Must know!
posted by segatakai to Technology (20 answers total)
It's somebody using a camera to film it from a TV or monitor. Note the station logo in the top-left.
posted by Pinback at 12:05 AM on June 4, 2012

(And I left off: What makes you think it isn't the case?)
posted by Pinback at 12:09 AM on June 4, 2012

I'm pretty sure this is stabilized video, to reduce the jerkiness of handheld filming. Compare. The image "box" moves and distorts so that the people and buildings inside the box don't shake around so much.
posted by dhartung at 12:12 AM on June 4, 2012 [7 favorites]

Yeah, I'm no expert, but this looks like an artifact of image stabilisation—the frame is being constantly repositioned so that the image itself doesn't move around.
posted by hot soup girl at 12:17 AM on June 4, 2012

(And I left off: What makes you think it isn't the case?)

The fact that the frame moves around but the screen image doesn't.
posted by phaedon at 12:23 AM on June 4, 2012 [2 favorites]

Response by poster: so in dhartung's link, I'm supposed to think the video on the right is "better"? Having the frame jump and dance is better than having the picture itself wiggle?
I ask again (in disbelief), is this a thing (that somebody thinks we all want)?

(I don't think it's a TV/monitor because you can't see *anything* (often you catch a glimpse of the monitor's frame, or a clock radio on top of the monitor, or *something*) in the black)
posted by segatakai at 1:17 AM on June 4, 2012

I agree - this is image stabilization. Without it, the video would be wildly jerking around so much you wouldn't be able to see a thing.

When the camera jerks down (or whatever direction), the software adjusts by moving the picture up - however, when it moves the picture up, there's no image data at the bottom of the screen to fill in the gap at the bottom, it is left black. Hence, the dancing black border.
posted by Salvor Hardin at 1:25 AM on June 4, 2012 [1 favorite]

I think the video is the result of someone using a camera that features image stabilization to film a monitor displaying footage shot by a handheld camera that does not have the image stabilization feature. Note how the logo in the corner jerks and jostles around in concert with the frame borders; if you were in the room watching the monitor the logo would be stationary and the image would be shaking around. The black bars on the left and right are the result of pillarboxing to make the 4:3 camera image fit a 16:9 frame, only the top and bottom bars are actually showing the edges of the monitor, which is the only light source in the room and thus seems to float in darkness.
posted by contraption at 1:27 AM on June 4, 2012 [1 favorite]

Most consumer video editing programs that have a stabilization function "zoom" in a preset amount to avoid the black borders. More advanced stabilization techniques use information from preceding and/or following frames to fill in the black borders, and/or generate picture elements that fill in the black borders using the same "smart fill" technique that the new Photoshop and others use to remove elements from still pictures. Usually you wouldn't even notice when it is done this most advanced (so far) way, unless you look closely.

Youtube now offers a stabilization function when a video is uploaded. I haven't tried that but I think it is the simplest type that gives those wobbly black borders. (Someone correct me if I am wrong about this on Youtube.)

I don't think that clip was filmed from a monitor - there would be different artifacts if it was. It looks like a direct copy of a video that was run through a stabilizer function, and perhaps the simple Youtube one (if I am right about Youtube).
posted by caclwmr4 at 1:34 AM on June 4, 2012

It's definitely a clip of someone filming a television. Youtube asks if you want to apply image stabilization when you upload a video, so they probably just clicked that.
posted by Rhomboid at 1:34 AM on June 4, 2012

The mismatch between the jumpy channel logo and the stabilized image it overlays, and the way the channel logo precisely tracks with the jumpy top and bottom borders, still lead me to believe that this is stabilized video shot from a stationary camera pointed at a monitor displaying shakycam footage from the disaster scene. Whether the stabilization was performed in-camera or by YouTube is tough to say, but it's definitely a strange effect.
posted by contraption at 1:53 AM on June 4, 2012

segatakai: "so in dhartung's link, I'm supposed to think the video on the right is "better"? Having the frame jump and dance is better than having the picture itself wiggle"

Yes. I'm interested in the picture, not the frame.
posted by IAmBroom at 5:37 AM on June 4, 2012 [1 favorite]

Yeah, definitely video stabilization. I recently made a little movie of a drive (taking video from the passenger's seat) on some fairly bumpy roads. When I uploaded it I decided to try YouTube's stabilization (I'd used it before on less shaky videos) but on this particular little movie I got the same result as your example. Immediately "de-stabilized". But it was amusing.
posted by mireille at 5:38 AM on June 4, 2012

Response by poster: ok, so I guess we can safely say this is some kind of image stabilization thing going on here; I appreciate everyone's input. I still have my doubts about what we are actually seeing here (video of a monitor or just stabilization software), but I'd like to do a bit of a piggyback on this question, if I may...

I am extremely flabbergasted that the jumping frame is considered an improvement on the jumping image, even after looking at a couple of examples I was able to find online, including dhartung"s link. I'm curious, am I alone in that regard?

If the outside of the frame got filled in, as calcwmr4 describes, I could imagine we might all agree that the end result was "better". But the dancing frame is, to my eyes, MUCH more distracting than a jerky video.

please weigh in, if you don't mind...
posted by segatakai at 6:56 AM on June 4, 2012

Yes, the dancing from is very distracting. That's a result of this particular flavor of image stabilization. It's a cheap, quick, and dirty and probably automated version. Other, more involved and hands-on versions would allow a stable crop to the results. This would, however, necessarily crop the image down, sometimes dramatically.

The source video used here is very likely a capture of some network broadcast.

I think the reason this particular video bugs you so much somewhat lies in the fact that this particular video doesn't really need image stabilization. Usually, image stabilization is used to gain stable imagery of an event that requires close examination. Say, for instance, the Zapruder film (which has, in fact, been re-processed using stabilization techniques)

Image stabilization necessarily results in a trade-off. You are willing to accept something like the dancing edges so that you can get a clear, unshaking video of what you really want to study. Again, this particular video doesn't really require stabilization. I agree that it was probably mistakenly applied in this case.
posted by Thorzdad at 7:13 AM on June 4, 2012 [1 favorite]

Typically you would crop the dancing frame. The image stabilization algorithm should be able to do this automatically by keeping track of the distortions it uses for each frame. I can't imagine someone would release software without a cropping option, so I guess the user chose not to crop on purpose.
posted by scose at 7:25 AM on June 4, 2012

On the high end, people who anticipate doing some image-stabilization try to shoot at a higher resolution than they're going to deliver, which allows them to crop out the edges of the frame without compromising quality. For instance, if you shoot wide angle with a 4K or 5K camera from Red, you've got quite a bit of latitude to reframe an image that's only going to be delivered at 1080p (which is a little smaller than a 2K image). In that case, you'd never realize there was image stabilization going on unless the stabilization was drastic and you had a really good eye — for instance, stabilization can match the positions of objects in the picture from frame to frame, but it can't (yet) get rid of the motion blur introduced in the original photography by a wildly moving camera.

What you're seeing here is definitely the result of a crude and automated image-stabilization process. I'd guess the user didn't crop this video because the frame dances so much that a large proportion of the original image would be lost.
posted by Joey Bagels at 7:32 AM on June 4, 2012

Compare the first hippo video on this page with the third.
posted by Rhomboid at 12:37 PM on June 4, 2012 [1 favorite]

Wow, what an excellent demonstration page, Rhomboid. Thanks and bookmarked. On that third video, the way the would-have-been black borders are filled from previous and following frames is most easily seen (to my eye) in the upper right. A few more plays will show it in the lower left and along the bottom. There's more that could be done until it is not noticeable at all.

It looks like Youtube uses the technique the writer of that page found was already discarded.
posted by caclwmr4 at 3:27 PM on June 4, 2012

I feel like maybe I'm not getting this across properly, so I'll try one more time to explain my interpretation which I feel quite certain is the correct one:

There was a dark room with a (probably widescreen) television in it. This television was displaying footage from the disaster scene, which was shaky and had no stabilization applied before broadcast. A person in the room watching the TV would see a stationary channel identification logo overlaid on a shaky handheld image.

Somebody with a stabilization-enabled digital camera pointed it at the screen while recording in a 4:3 aspect ratio and keeping the camera relatively stable. The camera's algorithm recognized that the majority of the image it was getting was shaking around and dutifully corrected it even though the shaking originated in the camera that shot the original footage, not in the camera running the algorithm.

This leads to a final video that has the contents of the screen (the original, shaky footage) stabilized while the space above and below the screen and the channel logo are treated as objects in the scene that were actually jostling around along with the camera.

Note that the black bars to the left and right are very distinct and do not move at all, and that the black borders on the top and bottom have blurry indistinct edges and are the ones that move around. These black borders haven't been automatically cropped away because they're not a byproduct of stabilization, they're the dark areas above and below the TV screen and are part of the image as far as the camera running the IS algorithm is concerned. You'd presumably see a similar effect if you used that camera to shoot video through the window of a moving car on a gravel road and kept the window frame in the image.

Note also that the channel logo moves around while the image behind it is stabilized. This indicates that stabilization was applied to the image on the screen by the camera filming it; if the camera filming the TV were introducing the shaking you'd expect to see that logo stabilized along with the video behind it.
posted by contraption at 10:13 AM on June 5, 2012

« Older Nancy Drew 10 years later   |   "Daily look" & closet organizer app for... Newer »
This thread is closed to new comments.