Space-Time Invariants for Recognizing 3D Motions from Arbitrary Viewpoints under Perspective Projection

摘要

Extracting visual motion is very important for understanding dynamic actions and for extracting dynamic events from video sequences. Recently, it was shown that some invariants on motions can be extracted from sequential images and applied for recognizing motions from images viewed from arbitrary viewpoints. Unfortunately, these space-time invariants were limited for planar motions viewed from affine cameras. In this paper, we propose a method for computing space-time invariants on non-planar motions viewed from two perspective cameras. The extracted invariants are applied for distinguishing 3D motions from video sequences viewed from arbitrary viewpoints.