Triangulation-Based Approaches to Three-Dimensional Scene Reconstruction - 3D Computer Vision: Efficient Methods and Applications - page 82

Fig. 1.27 0) in scene space along

with the adapted models, using the combined uniqueness and ordering constraint. ( a ) Scene 1
(fence with person). ( b ) Scene 2 (keyboard with hand). ( c ) Scene 3 (arm with bar). ( d ) Scene 4
(building with person)

Initial three-dimensional point clouds (corresponding to λ e =

Fig. 1.28 Initialisation ( top ) and final result ( bottom ) for scene 1 of the model-based stereo
method. For comparison, the three-dimensional point cloud illustrates the result of the fea-
ture-based spacetime stereo algorithm described in Sect. 1.5.2.5 with λ e 0, where the three-di-

mensional points belonging to the person are marked by a circle . The depicted initial configuration
corresponds to the highest offset from the final solution for which convergence of the model-based
stereo algorithm is achieved. This approach does not capture the object in front of the scene part
displaying repetitive structures

=