It may well happen, in some cases, that the observer cannot reconstruct exactly the object with the informations that he can get. The "visual hull" of the object is the best approximation of this object when the observer moves into this area.
We present an algorithm (based on an article by S. Petitjean) for computing visual hulls for a scene in 2D when the observation area encloses completely the object and is, in some sense, "far" from it. This algorithm has been implemented in C++ language using the LEDA library.
The report [French] is available (ps.gz, pdf, bib)
Reference: Sylvain Petitjean, A computational geometric approach to visual hulls, International Journal on Computational Geometry and Applications, 8(4):407-436, 1998.