Space-time invariants for recognizing 3D motions from arbitrary viewpoints under perspective projection

The visual motions are 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 these can be applied for recognizing motions from images viewed from arbitrary viewpoints. Unfortunately, invariants were limited for planar motions viewed from affine cameras. In this paper, we propose a method for computing invariants on non-planar motions viewed from perspective cameras. The extracted invariants are applied for distinguishing D motions from video sequences viewed from arbitrary viewpoints.