A semi-supervised approach to space carving

摘要: In this paper, we present a semi-supervised approach to space carving by casting the recovery of volumetric data from multiple views into an evidence combining setting. The method presented here is statistical in nature and employs, as starting point, manually obtained contour. By making use user-provided information, obtain probabilistic silhouettes all successive images. These provide prior distribution that then used compute probability voxel being carved. This setting allows us make background pixel information. As result, our combines advantages shape-from-silhouette techniques approaches. For process, propose new voxelated space. proposed projective one provides colour mapping for object voxels which consistent terms coverage with their projection onto image planes imagery under consideration. We quantitative results illustrate utility on real-world imagery.