Analysis of Two Methods for Estimation of Partial 3D Velocity

In this paper we analyse and compare two methods for the computation of the total 3D velocity along the optical axis of a stereo system, using the rigid body motion model. One of the methods uses the temporal and spatial derivatives of the depth information and the image flows and the other one uses the depth information as well as the differential flow between left and right images. Both methods are compared in terms of uncertainty propagation. The problems of generating ground truth data as well as the problems related to depth resolution are also discussed. Finally the results of our experiments to compute the 3D velocity in the optical axis using synthetic and real images are reported.