Shape from Shading
The two images below show the same surface under two different lighting conditions.  The image on the left was illuminated by with a parallel light source (as on a sunny day).   The image on the right was illuminated by a diffuse light source, as on a cloudy day.    

How can a vision system estimate the shape of such surfaces from either one of the images?   The usual way to approach this "shape from shading" problem is to assume that the lighting condition and the surface material are known, and then to estimate a 3D shape that can account for the given image intensities.    Many researchers had proposed solutions to this problem for images such as on the left, in which the light source is collimated as on a sunny day.   In my Ph.D. thesis,  I addressed problem for images such as shown on the right in which the light source is diffuse such as on a cloudy day.  

This shape-from-shading problem can also be addressed for human vision.    For example, you can "see for yourself" that, in the image on the left, the little isolated black square is sitting on a local hill, not in a valley.    Similarly, for the pair of little black squares in the image on the right,  the right dot is slightly higher on the surface than the left dot.   Although you might not be completely certain about such judgments and you might sometimes be wrong,  your visual system does remarkably well in making these judgements.    For more info, see the papers below and refs therein.

Selected Publications