Login | Register 
FEATURES
APPLICATIONS
DOWNLOADS
EDUCATION
BLOG
WIKI
FORUM
STORE

Eulerian Video Magnification

Helpful techniques for working with TouchDesigner. Answers only, no questions.

Eulerian Video Magnification

Postby ab30 » Fri Jul 20, 2012 3:52 am

An interesting technique for visualizing subtle changes in image sequences:

http://people.csail.mit.edu/mrub/vidmag/
ab30
 
Posts: 278
Joined: Fri Apr 27, 2012 10:36 am

Re: Eulerian Video Magnification

Postby matthewwachter » Fri Jul 20, 2012 12:52 pm

Cool! This is a great effect. I can think of how to do this with TOPs but it would be really cool to write a shader that does it.
matthewwachter
 
Posts: 151
Joined: Fri Jul 08, 2011 2:18 pm

Re: Eulerian Video Magnification

Postby matthewwachter » Thu Jul 26, 2012 4:19 pm

Here's what I ended up coming up with. It would be cool to figure out how to do the color amplification too.

motion_amplifier.zip
(1.89 MiB) Downloaded 241 times


Basically it does a subtract between the movie at the current time and the movie at (a) frames ago and gives you a monochromatic difference in motion between the two.

Then to calculate the direction in which the image is displaced, the monochromatic difference is sent through 4 different feedback loops in opposing x and y directions. Each chain of feedback is multiplied by the monochromatic difference at (b) frames ago thus giving you a positive brightness only if it travels in the direction of the feedback.

Then each feedback loop gets colored according to its appropriate displacement direction and finally, all of the feedback loops are composited into one signal, added over gray, and sent into a displacement top which displaces the original image (c) frames ago.

Hope this makes sense. Any suggestions of enhancements or modifications would be greatly appreciated. :D

Thanks!

-Matt
matthewwachter
 
Posts: 151
Joined: Fri Jul 08, 2011 2:18 pm

Re: Eulerian Video Magnification

Postby ab30 » Thu Aug 09, 2012 4:37 pm

Nice attempt Matthew.

At first sight, it seems that the resulting movement is a bit funky, but the main idea is here i think. Perhaps i would have seen some quaternions somewhere into the network though I didn't had time yet to really dig into the details of the original paper.That said, I keep it and your zip file as reference for future investigation. thanks for having given it a try ;)
ab30
 
Posts: 278
Joined: Fri Apr 27, 2012 10:36 am


Return to Techniques

Who is online

Users browsing this forum: No registered users and 1 guest