Loading...
Translational alignment, parametric motion, optical flow, and layered motion techniques.
Video adds a new dimension: time. Motion estimation extracts how pixels move between frames, enabling tracking, stabilization, and understanding of scene dynamics.
What is this chapter about? We learn to estimate motion at different granularities—from global camera motion to dense per-pixel optical flow. These techniques analyze how the visual world changes over time.
Why does this matter? Motion understanding enables:
How the topics connect: We start with optical flow—estimating motion for every pixel. Then video stabilization shows how to smooth out camera shake. Finally, we explore tracking algorithms that follow specific objects through time.
Click any topic to jump in
Per-pixel motion estimation — brightness constancy, Lucas-Kanade, Horn-Schunck, and coarse-to-fine pyramids.
Two applications of dense motion
Smoothing camera trajectories — motion filtering, rolling shutter correction, and crop trade-offs.
Synthesizing intermediate frames — motion-compensated warping, bidirectional flow, and neural methods.
Following targets through video — Kalman filtering, template correlation, and deep Siamese trackers.
This chapter is part of PixelBank Premium. Create a free account, then upgrade to read the full lesson — concepts, walkthroughs, and exercises.