Template-Free 3D Reconstruction of Poorly-Textured Nonrigid Surfaces
Two main classes of approaches have been studied to perform monocular nonrigid 3D reconstruction: Template-based methods and Non-rigid Structure from Motion techniques. While the first ones have been applied to reconstruct poorly-textured surfaces, they a
- PDF / 3,813,470 Bytes
- 16 Pages / 439.37 x 666.142 pts Page_size
- 24 Downloads / 208 Views
Xi’an Jiaotong University, Xi’an, China {wfx,zjz}@mail.xjtu.edu.cn, [email protected] 2 CVLab, EPFL, Zurich, Switzerland [email protected]
Abstract. Two main classes of approaches have been studied to perform monocular nonrigid 3D reconstruction: Template-based methods and Non-rigid Structure from Motion techniques. While the first ones have been applied to reconstruct poorly-textured surfaces, they assume the availability of a 3D shape model prior to reconstruction. By contrast, the second ones do not require such a shape template, but, instead, rely on points being tracked throughout a video sequence, and are thus illsuited to handle poorly-textured surfaces. In this paper, we introduce a template-free approach to reconstructing a poorly-textured, deformable surface. To this end, we leverage surface isometry and formulate 3D reconstruction as the joint problem of non-rigid image registration and depth estimation. Our experiments demonstrate that our approach yields much more accurate 3D reconstructions than state-of-the-art techniques. Keywords: Non-rigid 3D reconstruction Template-free shape estimation
1
·
Poorly-textured surfaces
·
Introduction
This paper tackles the problem of estimating the 3D shape of a poorly-textured, nonrigid surface in all the frames of a monocular video sequence. Reconstructing the 3D shape of a deformable surface from monocular images is a challenging task, which has attracted a lot of attention over the years. The resulting algorithms can be roughly classified into two categories: Template-based methods and Non-rigid Structure from Motion (NRSfM) techniques. Template-based methods [1–11] exploit the availability of a reference image in which the 3D shape is known and attempt to estimate the surface deformations in a new input image. To this end, they typically try to minimize an imagebased cost function, which encodes how well the deformed surface reprojects in the input image. Since this information alone leaves reconstruction ambiguities, existing approaches have developed various shape priors. In particular, Electronic supplementary material The online version of this chapter (doi:10. 1007/978-3-319-46478-7 40) contains supplementary material, which is available to authorized users. c Springer International Publishing AG 2016 B. Leibe et al. (Eds.): ECCV 2016, Part VII, LNCS 9911, pp. 648–663, 2016. DOI: 10.1007/978-3-319-46478-7 40
Template-Free 3D Reconstruction of Poorly-Textured Nonrigid Surfaces
649
Fig. 1. Our approach: template-free reconstruction of poorly-textured surfaces.
great progress has been made in template-based reconstruction by exploiting surface isometry [1–9], with methods yielding accurate reconstructions in an efficient manner [3,6,12,13] and, in some rare cases, even tackling the case of poorly-textured surfaces [1,2]. The main drawback of this approach, however, is its requirement for a 3D reference surface shape. Whether to model garments, mechanical structures, or human organs, one can in general not realistically expect having access to
Data Loading...
