/Eulerian_Motion_Magnification

This repository provides a python implementation of eulerian motion magnification.

Primary LanguagePython

Python Implementation of Eulerian Motion Magnification

This repository contains a python3 implementation of the paper "Eulerian Video Magnification for Revealing Subtle Changes in the World" .

Input Video Motion Magnified

Dependencies

  • scipy
  • scikit-image
  • opencv==3.4.1.15
  • numpy
  • pyrtools==1.0.0

We recommend installing the dependencies using conda.

Execution

We support two execution butterworth bandpass filtering and ideal bandpass filtering. For evaluation, we utilize the videos provided the authors here. Place the videos in the ./source folder.

We allow for command line arguments in each of the codes.

  • Lower cutoff frequency --fl
  • Upper cutoff frequency --fh
  • lambda_c --lambda_c ={2,3}
  • Number of pyramid levels --nLevels

Live Magnification

The Eulerian Motion Magnification code can also be used for real-time performance using only the intensity channel. We use butterworth bandpass filtering for real-time perfomance. Please run live_magnification.sh for real-time performance.

Citation

If you utilize this code for your work, please cite the following paper.

  author = {Hao-Yu Wu and Michael Rubinstein and Eugene Shih and John Guttag and Fr\'{e}do Durand and
  William T. Freeman},
  title = {Eulerian Video Magnification for Revealing Subtle Changes in the World},
  journal = {ACM Transactions on Graphics (Proc. SIGGRAPH 2012)},
  year = {2012},
  volume = {31},
  number = {4},
}