007, meet the "Motion Microscope" video processing tool...

[DevilishBooster]

If you watch the video, this is some pretty amazingly cool technology. Can you say "spy tech of the future"? 

TL;DR - It's video processing code that allows you to see "invisible motion" and can recreate audio from video of objects.

http://www.ted.com/talks/michael_rubinstein_see_invisible_motion_hear_silent_sounds_cool_creepy_we_can_t_decide?utm_source=feedburner&utm_medium=feed&utm_campaign=Feed%3A+TEDTalks_video+%28TEDTalks+Main+%28SD%29+-+Site%29

[Kingofpants]

The motion magnifying and colourshift magnifying is pretty cool. But the audio one seems impossible unless you have a crazy fast camera.

[Guest]

Well, the sound recovery requires crazy high frame rates and the recoveries they showed were from very detailed, up-close zoomed-in-on objects, so you're not going to be going around recovering sound from youtube or movies or whatever.

And speaking of youtube, I'm guessing heavy compression would make the colour-shift analysis unusable. You need high bitrate source to see tiny changes.

It is cool though, high frequency cameras are sure to become more affordable in the future. And props to them for making the code availabe to anyone.

[LeChef]

this is really cool

[Pesmerga]

That's quite awesome, I bet it could fuel someone's paranoia

[smjpl]

Wow this is amazing stuff. Turning video into a microphone is really spectacular.

 

Here is the main webpage (can be seen in TED vid) that contains links to publications (main paper is free) and the available source code (matlab): http://people.csail.mit.edu/mrub/vidmag/

[WanderingFool]

Well, the sound recovery requires crazy high frame rates and the recoveries they showed were from very detailed, up-close zoomed-in-on objects, so you're not going to be going around recovering sound from youtube or movies or whatever.

And speaking of youtube, I'm guessing heavy compression would make the colour-shift analysis unusable. You need high bitrate source to see tiny changes.

It is cool though, high frequency cameras are sure to become more affordable in the future. And props to them for making the code availabe to anyone.

 

The only thing about the high frequency camera's is the amount of frame-rate you actually need is pretty high.  Even assuming a lower toned female voice is 165hz, with nyquests theorem you could need at least sample at 330hz...which means a camera that can capture 330 fps.  If you go more of a higher ranged female talking voice of 255hz, you need 510fps on the camera.

[McMurderMonkey]

The only thing about the high frequency camera's is the amount of frame-rate you actually need is pretty high.  Even assuming a lower toned female voice is 165hz, with nyquests theorem you could need at least sample at 330hz...which means a camera that can capture 330 fps.  If you go more of a higher ranged female talking voice of 255hz, you need 510fps on the camera.

This is the problem with technologies so dependent on other technological advances.