Ohnishi Lab.
Computer Vision
Research Theme

| Overview | Brain Analysis | Computer Vision | Computer Audition | Natural Language | Rehab. Eng. |

| Top | Previous | Next |


3D Shape Reconstruction from Shaded Images

Two-image photometric stereo :
In the conventional photometric stereo method, the surface orientation of an object is determined by using multiple images. The multiple images are obtained by varying the position of light sources, and the surface orientation of the object can be determined only in the area of the object illuminated by all (at least three) light sources. We propose a method for determining surface orientation along with reflectance, in which only two light sources are used. It is assumed that an object is convex and has a smooth Lambertian surface with locally constant reflectance. We find a "separation line" in the image, along which the surface normal is represented as a linear combination of two vectors pointing in the direction of light sources. The "separation line" separates the surface into two regions. When the reflectance is known, of cause ; even in the case of reflectance is unknown, we can still determine the reflectance as well as the surface orientation of the object using property of the "separation line". Simulations and experiments carried out under various situations yielded satisfactory results.
Unique recovery of convex polyhedron :
We discuss the uniqueness of 3-D shapes recovery of a polyhedron from a single shaded image. We show that multiple convex (and/or concave) shape solutions usually exist for a polyhedron of which three or more facets meet at an apex (like a pyramid). It has discussed how to check and avoid the super strictness for a complex polyhedron. However, there is different problem for a simple polyhedron without super strictness. Horn has shown a numerical example in which two convex and two concave shape solutions exit for a trihedral corner. We describe that multiple shape solutions exit for a pyramid analytically. We also show the shape can be uniquely determined by using interreflections as a constraint to limit the shape solution for a concave polyhedron.



HOME RESEARCH PEOPLE JPN