VMV: Vision, Modeling, and Visualization

Permanent URI for this community

https://diglib.eg.org/handle/10.2312/1013

Browse

Browsing VMV: Vision, Modeling, and Visualization by Subject "and systems"

Now showing 1 - 4 of 4
  • Thumbnail Image
    Item
    3D Shape Matching based on Geodesic Distance Distributions
    (The Eurographics Association, 2012) Martinek, Michael; Ferstl, Matthias; Grosso, Roberto; Michael Goesele and Thorsten Grosch and Holger Theisel and Klaus Toennies and Bernhard Preim
    In this work, we present a signature for 3D shapes which is based on the distribution of geodesic distances. Our shape descriptor is invariant with respect to rotation and scaling as well as articulations of the object. It consists of shape histograms which reflect the geodesic distance distribution of randomly chosen pairs of surface points as well as the distribution of geodesic eccentricity and centricity. We show, that a combination of these shape histograms provides good discriminative power to find similar objects in 3D databases even if they are differently articulated. In order to improve the efficiency of the feature extraction, we employ a fast voxelization method and compute the geodesic distances on a boundary voxel representation of the objects.
  • Thumbnail Image
    Item
    Optimized Canonical Coordinate Frames for 3D Object Normalization
    (The Eurographics Association, 2012) Martinek, Michael; Grosso, Roberto; Greiner, Günther; Michael Goesele and Thorsten Grosch and Holger Theisel and Klaus Toennies and Bernhard Preim
    In this paper, we describe a method to optimize an orthogonal system of axes for 3D objects in order to perform normalization with respect to orientation and scale. An energy function evaluates the quality of a system by considering symmetry, rectilinearity and the origin of the system within the current axis aligned bounding box. Starting with the PCA-axes as initial system, we find a canonical coordinate frame by minimizing the energy in an efficient and elaborate optimization process. We provide a fully automatic normalization pipeline with the possibility to manually set various intuitive parameters in order to influence the outcome. The symmetry part of our energy function uses a combination of plane reflective and rotational symmetries. In this context, we introduce a novel continuous symmetry measure which is entirely implemented on the GPU. The high efficiency of the implementation enables us to find an optimal alignment for 3D objects interactively, making our method suitable even for large 3D databases. We also demonstrate the applicability of our framework for 3D shape matching by approximating the Hausdorff distance for 3D models.
  • Thumbnail Image
    Item
    Resolving Twisted Surfaces within an Iterative Refinement Surface Reconstruction Approach
    (The Eurographics Association, 2012) Annuth, Hendrik; Bohn, Christian-A.; Michael Goesele and Thorsten Grosch and Holger Theisel and Klaus Toennies and Bernhard Preim
    We present a method which resolves twisted surface regions within a surface reconstruction approach that uses local refinement operations to iteratively fit a surface into an unorganized point cloud. We show that this local operation can be integrated reliably and efficiently, although resolving twisted surfaces is not a local operation since it may cause modifications up to one half of the entire surface. We introduce a novel data structure called the minimal edge front that enables efficiently retrieving topological information from the surface under investigation. Equipped with this operation the algorithm is able to robustly handle huge point-clouds of complex closed and also not closed objects like landscapes.
  • Thumbnail Image
    Item
    SOAR: Stochastic Optimization for Affine global point set Registration
    (The Eurographics Association, 2014) Agus, Marco; Gobbetti, Enrico; Villanueva, Alberto Jaspe; Mura, Claudio; Pajarola, Renato; Jan Bender and Arjan Kuijper and Tatiana von Landesberger and Holger Theisel and Philipp Urban
    We introduce a stochastic algorithm for pairwise affine registration of partially overlapping 3D point clouds with unknown point correspondences. The algorithm recovers the globally optimal scale, rotation, and translation alignment parameters and is applicable in a variety of difficult settings, including very sparse, noisy, and outlierridden datasets that do not permit the computation of local descriptors. The technique is based on a stochastic approach for the global optimization of an alignment error function robust to noise and resistant to outliers. At each optimization step, it alternates between stochastically visiting a generalized BSP-tree representation of the current solution landscape to select a promising transformation, finding point-to-point correspondences using a GPU-accelerated technique, and incorporating new error values in the BSP tree. In contrast to previous work, instead of simply constructing the tree by guided random sampling, we exploit the problem structure through a low-cost local minimization process based on analytically solving absolute orientation problems using the current correspondences. We demonstrate the quality and performance of our method on a variety of large point sets with different scales, resolutions, and noise characteristics.