In this talk, we present a new geometry reconstruction algorithm from a
sequence of images of a rigid object. The main idea of the algorithm is
the following. For a moving object in a constantly illuminated
environment, the relative motion between the object and the illumination
source will produce a variation in intensity values, and this provides an
important cue for solving the reconstruction problem. In particular, it
allows us to compute the normal vector field of the object's surface, and
hence, the depth (3D structure) relative to the camera can be recovered by
integrating the normal vectors. The iterative algorithm we proposed is
quite simple and yet effective. Using some differential geometry, we will
discuss the algorithm's convergence in the second part of this talk.