Zombie thesis magnifies invisible motion in real time

What it does

Live Video Magnification is a Qt 6 desktop application that ingests live webcam feeds or video files and applies Eulerian magnification algorithms to expose tiny spatial and temporal variations. It can amplify periodic color changes—such as the cardiac cycle reddening a face—or tease out micro-motions invisible to the naked eye, all while letting you draw a region-of-interest box and export the results. The author calls it a “zombie project” born from a 2015 bachelor thesis that gets periodically reanimated with new build systems.

The interesting bit

The pipeline decomposes frames into spatial frequency pyramids, applies temporal bandpass filters to isolate specific motion wavelengths, amplifies those layers independently, and then collapses the pyramid back onto the original frame. This is classic MIT Eulerian Video Magnification packaged as a desktop GUI that can juggle multiple video sources at once.

Key highlights

• Supports simultaneous input from multiple cameras and video files.
• Two distinct modes: color magnification for periodic changes like pulse-driven skin flushing, and motion magnification with adjustable cutoff and chromatic attenuation.
• Region-of-interest selection to restrict processing to a specific area.
• Video export with selectable codecs.
• GPLv3 licensed, built on OpenCV and Qt6.

Caveats

• The README admits it is “more of a zombie project” that crashes and is only actively tested on Windows with Visual Studio 2022.
• Build instructions and environment assumptions are heavily Windows-centric; running elsewhere is undocumented.

Verdict

Worth a look if you need a ready-made GUI for experimenting with Eulerian video magnification or teaching signal-processing intuition. Skip it if you want a polished, actively maintained production tool.