Abstract Shadows in an image have been used for computing surface orientations in the corresponding scene. In scenes involving polyhedra, the surface reconstruction problem is underconstrained and, in general, requires that three appropriate parameters be specified in advance. Some cases need even more information. Multiple sources of illumination can be used to advantage under these situations. It has been shown in this paper that such sources permit reconstruction with just the three pieces of information cited above. Shadow geometry alone is, however, not powerful enough to effectively solve the scence understanding problem if the scene has objects with curved surfaces. Integration of the knowledge implicit in the texture (if any) on the surface with that provided by shadows is proposed as a way out of this situation. An effective technique has been given for computing the standard texture elements with the aid of shadows present in the image. This, it will be seen, is the key to the indicated integration.