Some remarks on the shape-from-shading problem in computer vision☆