This paper studies the determination of the rigid body transformation allowing the automatic correlation of the 3-D information inherent in different imaging devices. It addresses a part of the overall methodology, namely the global registration problem (e.g translation and rotation estimation). Several approaches have been reported according to the representations of geometric data (point and line segment sets, curves, surfaces, volumes). Two original schemes have been derived when the registration operates directly on the 3-D sampled volume. The first one is based on 2-D registration of multiple 2-D binary projections of the objects. The second one makes use of quaternion theory.