Photometric stereo is a method to reconstruct an object’s surface using pictures of the object taken from different directions. Applications of photometric stereo include face recognition, analysis of astronomical bodies’ surfaces [1], medical imaging [2], and dermatology [3].
This paper presents a new photometric stereo technique to recover 3D surfaces. The main contribution of this approach is that it eliminates the restriction that the reflections should be diffuse; therefore, the authors are able to reconstruct specular surfaces. In addition, this method allows arbitrary positions of lights, any kind of light attenuation, and can be implemented by fast marching.
The authors demonstrate the effectiveness of their approach, comparing it with other techniques in the literature. The difference is remarkable because the other techniques do not handle specular reflections or arbitrary light sources. Finally, using a camera, a toy figure, and a coin, it is shown that the new model can reconstruct real-world objects reasonably.