Silhouette and stereo fusion for 3D object modeling

Abstract

In this paper we present a new approach to high quality 3D object reconstruction. Starting from a calibrated sequence of color images, the algorithm is able to reconstruct both the 3D geometry and the texture.

Model-based human shape reconstruction from multiple views

“…Objectcentred reconstruction has been used to improve reconstructed scene models of people in model-enhanced stereo [13] where an initial surface reconstruc-tion is used to restrict the search range to refine stereo correspondence. More recently techniques have been presented to refine a robust initial estimate of shape derived from silhouettes by incorporating stereo and photo-consistency from multiple views [11,16,17]. These techniques combine shape cues for robust reconstruction without the requirement for a prior surface model but provide no structure for animation synthesis.…”

Variational Shape and Reflectance Estimation Under Changing Light and Viewpoints

“…When the estimated surface is far from the actual surface widely separated views sample from different regions of the underlying surface, implying that the gradient may not provide meaningful information. In this case, it is helpful to restrict the number of views used in g. For choosing a subset of the observations, we select the n cameras closest cameras to the median visible camera [5], where the middle camera is currently based solely on the azimuthal angle. This modification fits into our framework by changing the visibility function into V , which returns 0 for all cameras that are not in the considered range.…”

“…That works well for textured Lambertian surfaces but fails in the presence of specular highlights or uniform texture. Additionally, stereo-based techniques reconstruct only the shape and not the surface reflectance properties even though some approaches can handle specular objects using robust scores [4,5].…”

“…Most of the approaches either filter or remove specularities and use only diffuse observations in the reconstruction [5]. Another option is to design similarity scores that account for specular highlights either using constraints on the light and camera positioning [10] (Helmholtz reciprocity) or on surface material [11] (uniform material under constant light).…”

Continuous Global Optimization in Multiview 3D Reconstruction

“…Generally, there are two types of techniques to combine silhouette information and photoconsistency. The first strategy integrates silhouette constraints into stereo-based optimization [11,22,25]. The alternative is to use the visual hull merely as initialization for a stereo-based technique [26].…”