Toward surface reconstruction using multi-image shape from shading

verfasst von
Christian Heipke, Christian Piechullek
Abstract

We investigate multi-image shape from shading. Using this method the heights and the albedo of the object surface are simultaneously estimated from the pixel grey values of any given number of images taken from arbitrary camera positions, and control information. The perspective projection is used for the transformation from object into image space. The existence and the uniqueness of the solution of the described algorithm is analyzed. The following results have been derived: (1) One image allows for the reconstruction of surface inclinations of a surface assumed to be a bilinear mesh. (2) Two images with different exterior orientation allow for the calculation of surface heights. It is not necessary to use conjugate points. (3) If at least three images with different light source positions are available, the reconstruction of arbitrary surfaces with varying albedo is possible. For selected cases the results are verified using synthetic images. The theoretical standard deviations of the unknown heights are used to assess the quality of the surface reconstruction.

Externe Organisation(en)
Technische Universität München (TUM)
Typ
Aufsatz in Konferenzband
Seiten
361-369
Anzahl der Seiten
9
Publikationsdatum
1994
Publikationsstatus
Veröffentlicht
Peer-reviewed
Ja
ASJC Scopus Sachgebiete
Elektronische, optische und magnetische Materialien, Physik der kondensierten Materie, Angewandte Informatik, Angewandte Mathematik, Elektrotechnik und Elektronik