< Previous | Contents | Manuals Home | Boris FX | Next >
How It Works
The system analyzes multiple frames at once to estimate how image content moves over time.
It calculates a dense motion field, meaning a motion vector is generated for every pixel (not just specific features).
These motion fields are used to help track and link features frame-to-frame, supporting more robust tracking in cases where traditional methods may drop out.
Internally, it uses a hierarchical, multi-scale approach: footage is analyzed at successively lower resolutions to simplify motion estimation, then refined at higher resolutions.
The model also recognizes higher-level visual patterns, like edges, shapes, or semantic regions, which can help maintain tracking in low-texture or blurred areas.
©2025 Boris FX, Inc. — UNOFFICIAL — Converted from original PDF.