Reconstruction of Complete 3D Models by Voxel Integration

Abstract

A method for reconstructing a complete three-dimensional (3D) voxel model from multiple images captured during two circular motions of an object is proposed. The two circular motions include the object with a regular orientation and the corresponding reversed orientation. The shape-from-silhouette method, which constructs a voxel-based 3D model from the silhouettes of two-dimensional (2D) images, is used for the reconstruction. In general, a single camera focused on an object's circular motion cannot capture the entire object using a single fixed orientation. Thus, it is required to invert the orientation of the object at least once to capture the entire object. In this paper, a novel method for integrating two incomplete reconstructed voxel models to produce a single complete voxel model is discussed. Complex examples are provided to demonstrate the effectiveness of the proposed algorithms.

Keywords

• reconstruction
• shape from silhouette
• voxel integration
• spin-image
• icp