Case Study: MIT CSAIL achieves visualization of imperceptible motion and color changes with MathWorks MATLAB

MIT CSAIL Researchers Develop Video Processing Algorithms to Magnify Minute Movements and Changes in Color

MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) researchers developed video‑magnification algorithms to reveal imperceptible movements and tiny color changes—such as measuring heartbeats from ordinary video—but faced the challenge that cameras record only extremely small intensity variations that require careful spatial and temporal filtering, faster execution, and easy code sharing. To build and prototype their methods they relied on MathWorks’ MATLAB and its toolboxes.

Using MathWorks’ MATLAB plus Image Processing Toolbox, DSP System Toolbox, Parallel Computing Toolbox and MATLAB Compiler, the team implemented spatial decomposition and temporal filtering, sped execution by processing multiple frames and experiments in parallel (e.g., on a 24‑core system using parfor), and compiled standalone executables for Windows, Linux, and Mac. The outcome was measurable: subtle signals (pulse and vibration‑derived sound) became visible, collaboration improved because MATLAB code was easy to share and read, experiments ran much faster in parallel, and the algorithms were broadly distributable to users without MATLAB.

MIT

Michael Rubinstein

CSAIL Research Affiliate