Revealing Subtle Heat Flows
all around us
using Microbolometric Videos

Mani Ramanagopal· Akihiko Oharazawa· Sriram Narayanan· Zeqing Yuan· Srinivasa Narasimhan

Carnegie Mellon University
🔔

Coming soon — Paper, code, and data will be released soon.

TL;DR — Low-cost thermal cameras have a reputation for being noisy and are presumed to be insufficient to capture subtle heat transport. We upend this conventional wisdom by a simple change in how we process microbolometric videos. Based on the physics of image formation in microbolometers, we show that first applying per-pixel 1D temporal transform (using a shift-invariant, linear phase, compact support filter) and then spatially denoising in the coefficient domain enables extracting rich spatiotemporal heat flows with existing low-cost hardware.

Experimental Results 🔥 more scenes cooking

Scene 1 Bunny Wind Draft
Temporal Pixel Plot
Scene 2 Yoda's Eyes
Temporal Pixel Plot
Scene 3 Book & Keyboard
Temporal Pixel Plot
Scene 4 Vehicle Exhaust
Temporal Pixel Plot
Scene 5 Spinning Fan
Temporal Pixel Plot
Scene 6 Rotating Polarizer
Temporal Pixel Plot
Scene 7 Bandpass LWIR Filters
Temporal Pixel Plot

Poster