Quality improvement of 3D models reconstructed from silhouettes of multiple images

Shape-from-Silhouette (SFS) is a technique for estimating the shape of an object from images of its silhouette. It is a simple but effective technique for obtaining a three-dimensional (3D) model of an object by using two-dimensional (2D) images of the object in different views. The inherent problem with this method is visual features or irregularities, such as sharp edges, corners, and artifacts. These affect not only the outline shape of a model, but also the accuracy of texture mapping, which is essential for creating a 3D color model. The objective of this study is to propose a quality improvement method to address the aforementioned problems for the SFS method. The proposed quality improvement method is a multiple-step iteration procedure combining a re-meshing process and a mesh smoothing process to remove virtual features and artifacts on the original model, while preserving the image silhouettes and the smoothness of the model. Two algorithms based on regular and irregular meshes, respectively, are described and compared. Several examples are presented to verify the feasibility of the proposed algorithms.