Although photometric stereo allows direct shape recovery form two images, we can infer a lot about shape from a single image. To recover the shape of a surface or object from shading information in a single image we shall proceed in two stages. First we assume that we know the surface orientation at a particular point (or points). In general these are either maxima in the reflectance function where, from Equation 3:

the surface normal and source are coincident, or occluding boundaries, where the surface normal is orthogonal to the viewing direction and the corresponding point on the contour. Second we attempt an iterative solution, known as characteristic strip expansion [Horn, 1986]. This grows a curve in space on which the orientation is known from the initial seed point or points; it is by no means the only method but serves to illustrate the general principle.