3D model reconstruction

Abstract

This project aims to reconstruct and improve 3D object models by using several methods. First, images of the object at various viewpoints are captured by using camera. The color images are converted to the silhouette images. The object silhouette is extracted from the images. The 3D object model is reconstructed from the object silhouettes which contain object shape information. The method of 3D reconstruction is called shape from silhouettes (SFS). Then, the marching cubes algorithm is used to generate the 3D object polygonal model. It subdivides the 3D space into small cubes. The algorithm will scan through each of the cubes and replacing the cube with an appropriate set of triangle. However, the concave surface of the object may not be reconstructed accurately. So, the Helmholtz Reciprocity principle is applied on the 3D object model. The shape of surfaces with arbitrary bidirectional reflectance distribution function (BRDF) can be recovered. Finally, the 3D object model with both convex and concave surfaces is reconstructed.