Abstract

An important subject in computer vision is to determine 3D scene structure and motion from multiple 2D images taken by one or more, stationary or moving cameras. When there is no relative motion among the objects in view and the camera, and only one fixed camera is available, one tries to extract 3D information using techniques such as shape from shading, or shape from texture. When there is no relative motion, but more than one camera is available, shape from stereo is applicable. Suppose there is relative motion (which may be due to the motion of one or more objects, the motion of the camera, or both), 3D information can be obtained by the techniques generally referred to as shape from motion. If there is motion and more than one camera, one can integrate the information from both the stereo module and the motion module, e.g., [Mit88], [Wax88], Note that stereo problem can be treated as a special case of motion if the spatial distribution of the camera is considered as the result of "instantaneous" camera movement.

© 1989 Optical Society of America