Abstract

This paper presents an algorithm for voxel-based reconstruc- tion of ob jects with general refiectance properties from multiple calibra- ted views. It is assumed that one or more reference ob jects with known geometry are imaged under the same lighting and camera conditions as the ob ject being reconstructed. The unknown ob ject is reconstructed using a radiance basis inferred from the reference ob jects. Each view may have arbitrary, unknown distant lighting. If the lighting is calibra- ted, our model also takes into account shadows that the ob ject casts upon itself. To our knowledge, this is the first stereo method to handle general, unknown, spatially-varying BRDFs under possibly varying, di- stant lighting, and shadows. We demonstrate our algorithm by recover- ing geometry and surface normals for ob jects with both uniform and spatially-varying BRDFs. The normals reveal fine-scale surface detail, allowing much richer renderings than the voxel geometry alone.