A semisupervised approach to space carving

Abstract

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

The importance of silhouette optimization in 3D shape reconstruction system from multiple object scenes

“…In addition, a model for the object surface reflectance is also required [21]. Space carving technique is use to allocate voxels to a 3D object by using the photo-consistency of a specific point with all its matching pixels from the given set of images and carves the 3D object from its background [17,18]. The voxel model produced by space carving is invariably noisy and limited to the visibility of surrounding objects that can be merge into the final 3D shape even they are separated in other views [22].…”

“…In order to avoid the limitations of data-driven and multi-view stereo approaches, volumetric scene modelling methods: shape from photo-consistency, shape from shading or motion and shape from silhouettes are proposed in [16,17,18,19,20]. The photo-consistency methods requires camera parameters for each view, colour consistency constraints and criteria for consistency checker with specific threshold or from user input.…”