Hacker News new | past | comments | ask | show | jobs | submit login
Rolling Shutters (jasmcole.com)
708 points by hazz on Oct 13, 2014 | hide | past | favorite | 41 comments



Nice! Whenever I see rolling shutter photos on flickr/etc I always think about this old page where a fellow built a long-distance camera from a flatbed scanner to get the effect intentionally: http://www.sentex.net/~mwandel/tech/scanner.html

(There's a great image of a garage door opening & closing about 2/3 of the way down the page if you don't feel like reading the whole thing.)


This is so fascinating. This makes me long for the days when browsing somebody's homepage on the web felt like they were actually inviting you into their home like a friend.


His woodworking videos are fantastic. Also, FWIW, he's now got a whole page about the cameras he's used to film them:

http://woodgears.ca/misc/best_camera.html


You bet. Sometimes I feel a lot of these homepages were lost to page ranks and our quest to search everything on Google.


Same here. I love reading the pages of these super-creative types of people.


I did the same, as many other did:

http://camillomiller.com/sc

That website was of some inspiration, great too see it's still online!


Fellow scannercam hobbyist checking in! Did you also discover the trick of taking "HDR" photographs by setting the software to do a "colour" scan with every colour channel set to a different exposure?

Anyway, in my case the hobby led me down the rabbit hole of slitscanning and other spacetime-warping artworks, such as Bill Spinhoven's "It's About Time" (who was one of my tutors in art school, and AFAIK was the first artist to do this in a real-time installation)

http://www.spinhoven.nl/work/EN/workitsabouttime2.htm

These day's it's quite easy to do with software. Here's a test of one of my own prototypes made with Processing:

https://www.youtube.com/watch?v=S_7uWAUIjQU


I went down this rabbit hole myself, it's great fun. On our camera we noticed The RGB sensor data lined up well moving right-to-left, but misaligned when scanning left-to-right, creating a subtle rainbow effect.

http://topologo.com/animations/scan.html


Wow, that's great! Thanks for the links!


Cool!

It reminds me of the slitscan special effects technique that was used to create the stargate sequence at the end of 2001: A Space Odyssey.

http://filmmakeriq.com/lessons/slit-scan-recreating-the-star...


Conveniently available as a really cool iPhone app:

http://funnerlabs.com/apps/slitscan

I'm a big fan of this because (a) it gives you a photo with a time dimension! and (b) it happens to be a similar technique to what some satellites use (a single row of pixels that sweeps across the sky) so it's cool to gain an intuition into how that works.


It's a neat analysis from a mathematical perspective, but (especially for a rotating component like this) wouldn't the lighting be all wrong for the remapped pixels? The slow-speed scanning examples use a fixed image (note the highlight doesn't change) so it's likely not usable for real-world digital photography without updates to account for lighting.


Not only lighting would be wrong, but there would be holes in places where blades were before remapping.


Why would it be wrong? The pixels have the correct lighting (assuming the overall ambient light didn't change during the rotation), just the wrong position.


They have the correct lighting at their current position (the propeller was actually at that point in the image when the photo was taken). Which means when you transform them you'll get wrong lighting in the severely bent parts because they rotated elsewhere when the picture was taken. Light shines on it from another side and you cannot fix that just by changing the image transformation. It's not the ambient light, but the directional part.


A yes, the directional light (highlights, specs etc) would change...


Ha, my friend used the same photo as the example for his mathematical analysis of the rolling shutter effect: http://danielwalsh.tumblr.com/post/54400376441/playing-detec...

The questions he investigated: "Can we figure out the rate at which a propellor is spinning by analyzing this kind of photo? And can we figure out the real number of propellor blades in the photo?"


You can play a more complex version of that game using video of the vibrating strings on a musical instrument. Here's an especially good example: http://vimeo.com/4041788


Sony is making steady advancements in the global shutter with CMOS sensors. A bit harder on DSLRS with larger sensors and more pixels to read but the smaller sensors with smaller megapixels already have them [1]. So it's matter of a time that most CMOS bases videos will be free of rolling shutter, starting with higher-end video cameras that have sensors with just enough pixels to cover 2k-4k videos [2]

[1] http://www.sony.net/Products/SC-HP/new_pro/december_2013/imx...

[2] http://www.newsshooter.com/2014/09/11/io-industries-4k-super...


Couldn't help posting this (uses a Sony global shutter sensor):

http://scolton.blogspot.co.uk/2014/05/grasshopper3-mobile-se...

If it were possible to get this in a consumer grade video product I would be very happy. Unfortunately these sensors are $1295.


GLobal shutter is one of the reason why Sony and the like trailed RED in the high res film camera game.

There are a few cameras out there with global shutter, RED isn't one of them.

That and the colour reproduction is one of the many reasons why the arri alexa is popular despite have "less resolution"


Somewhat relatedly, check out this awesome new camera technology which essentially captures a rolling diff of the image rather than the image itself, with impressive results: https://www.youtube.com/watch?v=LauQ6LWTkxM


I see this effect happening in skydiving videos quite often.

The rolling shutter is also why stills from gopro videos never quite live up to how clear the videos look in motion.

The cover photo from this month's parachutist magazine is a great example:

http://parachutistonline.com/sites/all/files/images/cover201...

Notice the right leg of the jumpsuit, its flapping in the wind as the shutter rolls over the scene.

When people use the slow-mo feature for gopro videos everything kind of morphs rather than moving naturally. I've always found it to be a cool effect:

https://www.youtube.com/watch?v=dUSF6xmmqJg&t=46s


> The rolling shutter is also why stills from gopro videos never quite live up to how clear the videos look in motion.

I've thought about this as well. I always assumed the lack of clarity of single frames extracted from video material is because the eye/brain incorporates several images shown really quick in succession into one whole image. So when only a single frame is shown, there's not as much information in that as, say, 10 frames shown quickly in succession.


That's also because of the 'shutter speed' or 'shutter angle'. Basically each frame stays 'open' for a certain amount of time to let the light in. Anything slower than say 1/500s is going to introduce motion blur into the frame.

Shutter Speed/Angle are also creative choices. Anything that is faster than 1/50th (180 degree shutter angle) will introduce a 'staccato' effect to the video. Anything slower will be mushy/blurry.

Somewhat off topic, but interesting nontheless!


A very cool article, indeed; but I believe he uses the term exposure wrong.

Exposure is the total time our whole light sensitive area is exposed to the light coming from our scene. You can think of it as an integral of the sensor (or film) area exposed as a function of the time, divided by the total sensor area.

In the examples he uses the term exposure to describe the total scantime of the sensor, whilst it seems that his actual exposure (which is equal to the time each row of pixels samples the scene) is much smaller.

It may sound as a small difference but if one wants to reproduce the effect, we will essentially need to match two parameters: exposure and scantime. While exposure is easy to set, scantime is pretty much hardcoded and depends on the physical characteristics of the camera. Even an analog shutter has a scantime on small exposure times.


If I understand this correctly, it is effectively doing what a photo-finish camera does at race sports events, except that the slit moves across the scene, rather than the scene moving past the slit.

Photo-finish shots also end up looking pretty weird: http://coachdeanhebert.files.wordpress.com/2007/08/100-photo...


And the performance of those photo-finish systems is impressive. I would be grateful if someone could explain how the software is able to almost instantaneously identify the runners who are frequently not in lanes and also often have missing numbers. For one example, see FinishLynx (http://www.finishlynx.com/)


I can see it now. Soon Adobe will include some tool or setting in Photoshop that will automagically "fix" rolling shutter.


There are two categories of artifacts caused by rolling shutter: those from objects moving in the scene (such as the propeller in this example), and those that result from movement of the camera relative to the scene (especially pans, tilts, dolly movements, etc).

Those caused by camera movement (often resulting in an image that looks skewed) are somewhat easier to "fix" as you say with post processing since a correcting transformation can sometimes be applied to the entire image uniformly. Existing tools can do this with some success, but there are still some camera moves that prove to be more difficult (zooms, irregular movement, etc).

Artifacts caused by objects moving in the scene are often much more difficult to remove, at least when it comes to providing a generic solution, because "reconstruction" of the image requires fairly accurate information as to how those objects were moving. In the case of the rotating prop or wheel, it may be somewhat simple (the algorithm would still probably require user input of things like the center of rotation and speed), but in other cases, the motion may be quite complex (e.g., multiple wheels rotating in different directions/speeds, linear vs angular motion, etc). And that doesn't even account for the fact that in most cases, there will be occlusion in the source image from artifacts of the rolling shutter. That is, in the propeller example, you have patches of the background that are covered up in the source, but wouldn't be in a "fixed" image, so they need to be filled somehow.

What I'm saying is that sophisticated software may be able to do a lot in helping to correct for rolling shutter artifacts, but I don't think there will be an automatic, fix-all solution from post processing software any time soon.


For 2D-like pictures (legs of fan pictured from bottom), the fix might be possible. For 3D (propeller blades with every blade having a right and a left face), it will be not.

The software will need to choose the final position of blades in the rendered picture. There is no ideal position since the blades have been moving throughout the scan time. Whichever position we decide, there will be information missing for one/more blade. Say, left face of blade no. X needs to be rendered but the camera only captured its right face. May be assuming that all blades have same shape and information of one blade can be used in rendering other blade will fix the problem.

Also missing background will have to be reconstructed. That's another issue.


I seem to recall that Adobe Premiere Pro and/or Adobe After Effects contain a tool for mitigating rolling shutter artifacts, though I am not certain if it uses the method described in the article.


YouTube's stabilisation option attempts to remove (with reasonable success, in my experience) rolling shutter wobble caused by camera movement, although I don't believe it will mitigate the effect of a moving subject.

There was a talk at last year's Google I/O with some examples:

http://youtu.be/QdXugkXBTbc?t=15m56s


There is already a tool bundled with Nuke. I think its now in after effects (I'm pretty sure its the same tech.)

http://www.youtube.com/watch?v=Zt0u9hsPuZY


Awesome animated GIFs, definitely helps explain the concept.

This effect was manipulated to extract more information for this: http://newsoffice.mit.edu/2014/algorithm-recovers-speech-fro...


Definitely check out other articles on the author's blog; he's a great technical writer.


My favorite rolling-shutter video, of an upright bass: http://vimeo.com/4041788


The radial graph half way down the page reminds me of this: http://en.wikipedia.org/wiki/Ulam_spiral#mediaviewer/File:Sa...


Well that's certainly interesting. I was about to say, it reminds me strongly of a zeta function, in a way. But it turns out Dedekind's eta function was what I was thinking of [1]

[1] https://en.wikipedia.org/wiki/Dedekind_eta_function#mediavie...


Rolling shutters were also used by traditional cameras. This effect is really old school stuff. Rolling shutter providers better exposure than circular shutter. I remember that most of professional photographs taken in 80s also used rolling shutter.


You mean curtain shutters, which become just a slit traveling across the sensor at short exposures? I am pretty sure modern DSLRs still use them.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: