Multi-view 3D model reconstruction: exploitation of color homogeneity in voxel mask

This project aims to reconstruct a 3D object model from an image sequence captured by a mobile camera. The "shape-from-silhouette/photo-consistency" approach is adopted for the volumetric model reconstruction. To correct the concavity error, a novel voxel mask in 3D space for measuring the color dispersion, instead of using the conventional block matching technique, is proposed to locate the concavity of the object surface. In addition, a new technique for partial surface estimation is implemented to overcome the reconstruction problem due to insufficient texture information or color homogeneity. Finally, a single texture map, which is generated from the original multi-view images, is rendered to the surface model.