Volume intersection algorithm with changing directions of views

An algorithm is proposed for reconstructing a three-dimensional object from multiple silhouettes. The silhouettes are represented as binary arrays, and the three-dimensional object is represented as a linear octree. The silhouettes are taken in any viewing direction lying on one of the three coordinate planes with reference to the front face of the object. The authors propose a method of choosing automatically a direction of view using the information from the current reconstruction. Thus, the proposed method is more advantageous than earlier approaches where the silhouettes are obtained for a set of prespecified directions independent of the objects. The authors gradually generate the geometrical properties of the object during the reconstruction and, hence, determine a direction so as to get a better reconstruction. A method to quantify the degree of accuracy is proposed