Volume intersection is one of the simplest paradigms for recovering shape from 2D images. The underlying principle in this method is reconstructing an object by intersecting the volumes obtained by backprojecting its silhouettes in their corresponding directions. The set of silhouettes could be predetermined or it could be decided dynamically for optimum reconstruction. The former approach is passive while the latter comes under the purview of active vision. In this paper, the author attempts to integrate his research results in obtaining a linear octree description of an object from its silhouettes making use of a data structure called range[i,j,O/1]. The problem of shape from silhouette is formalized and it is shown here that the conventional approach of volume intersection for this problem need not be always efficient. The advantage of active vision technique is also discussed in the present context.