A hybrid approach to structure-from-motion (abstract only)

Publisher Website
Google Scholar
Abstract
A method is presented for computing structure from the motion of rigid objects which are rotating about a fixed axis. The input consists of two discrete frames containing the positions and instantaneous direction vectors of three points in orthographic projection. Because only the direction of the velocity vectors and not their magnitudes is needed, the method is insensitive to errors in velocity magnitude estimation. This type of computation could be important in recovering the 3-dimensional structure of objects under dynamic viewing conditions because viewer motion about stationary objects will generate fixed axis rotations.

This publication has 0 references indexed in Scilit: