Abstract

We present a method – termed Helmholtz stereopsis – for reconstructing the geometry of ob jects from a collection of images. Un- like most existing methods for surface reconstruction (e.g., stereo vision, structure from motion, photometric stereo), Helmholtz stereopsis makes no assumptions about the nature of the bidirectional reflectance distri- bution functions (BRDFs) of ob jects. This new method of multinocular stereopsis exploits Helmholtz reciprocity by choosing pairs of light source and camera positions that guarantee that the ratio of the emitted radi- ance to the incident irradiance is the same for corresponding points in the two images. The method provides direct estimates of both depth and ?eld of surface normals, and consequently weds the advantages of both conventional and photometric stereopsis. Results from our implementa- tions lend empirical support to our technique.