The forensic analyst is often faced with a single image of an unknown scene taken with an unknown camera. Under these conditions, it is difficult, if not impossible, to get useful information; however, using the procedure described in this paper, a 3D reconstruction of the scene is provided by analytical and geometric methods, some of which incorporate novel and robust approaches. In addition, a software package ("sv3DVision") has been developed that provides a flexible and easy-to-use environment, even by practitioners who are not experts in photogrammetry. One section of this paper describes the geomatic approaches dedicated for 3D crime-scene recording and reconstruction. This is followed by a section that explains in detail the full photogrammetric process developed by using a single image. Another section reports experimental results applied to simulated crime scenes. The concluding section notes where the proposed approach is successful. The authors advise that the method is applicable with scenes that have strong geometric contents. It is not valid for irregular and complex shapes with concave, convex, or curved forms. In addition, the images acquired must be oblique, having three vanishing points. Finally, in order to overcome the image-based modeling problem, the user must know some information about the scene, for example, a distance along with some geometric constraints. 9 tables, 9 figures, and 31 references