3DPVT2010 e Proceedings

Embed Size (px)

Citation preview

  • 8/3/2019 3DPVT2010 e Proceedings

    1/17

    May 20 (Thursday) - 10:15 - 11:15

    Poster Session 2

    10:15 - 11:15 - May 20 (Thursday)Session chair: Bob Fisher, University of Edinburgh

    3D Photography and Image-based Rendering

    Improving the accuracy of fast dense stereocorrespondence algorithms by enforcing local consistency of disparity fields

    Stefano Mattoccia

    UNIVERSITYOFBOLOGNA DEIS-ARCES, ITALY

    Accurate, dense 3D reconstruction is an important requirement in many

    applications, and stereo represents a viable alternative to active sensors.However, top-ranked stereo algorithms rely on iterative 2D disparityoptimization methods for energy minimization that are not well suited to the

    fast and/or hardware implementation often required in practice. Anexception is represented by the approaches that perform disparityoptimization in one dimension (1D) by means of scanline optimization (SO)or dynamic programming (DP). Recent SO/DP-based approaches aim toavoid the well known streaking effect by enforcing vertical consistency

    between scanlines deploying aggregated costs, aggregating multiplescanlines, or performing energy minimization on a tree. In this paper we

    3DPVT2010 e-proceedings http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed...

    117 2012-2-17 05:13

    http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceedhttp://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed
  • 8/3/2019 3DPVT2010 e Proceedings

    2/17

    Presented by Stefano Mattoccia

    May 20 (Thursday) - 10:15 - 11:15

    show that the accuracy of two fast SO/DP-based approaches can bedramatically improved by exploiting a non-iterative methodology that, bymodeling the coherence within neighboring points, enforces the localconsistency of disparity fields. Our proposal allows us to obtain top-ranked

    results on the standard Middlebury dataset and, thanks to its computationalstructure and its reduced memory requirements, is potentially suited to fast

    and/or hardware implementations.

    Planar patch detection for disparity maps

    Eric Bughin1, Andrs Almansa2

    1CMLA - ENS CACHAN, FRANCE, 2ENST TELECOMPARISTECH, FRANCE

    We propose a new parameter-free method for detecting planar patches indisparity maps. We first introduce an a contrario decision criterion which may

    be used to solve two decision problems on configurations of 3D points: (i) isthe configuration well explained by a plane?; (ii) what is the optimal number

    of planes that best explains the configuration? These decision criteria are thecore of an algorithm that searches for an optimal explanation of a disparity

    map by planar patches whenever applicable. This method may be used for 3Dreconstruction of urban environments, particularly in the context oflow-baseline stereo where precision requirements are most strict, and apertinent choice of the type and amount of regularization is key to achieving

    accurate results.

    3DPVT2010 e-proceedings http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed...

    217 2012-2-17 05:13

    http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceedhttp://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed
  • 8/3/2019 3DPVT2010 e Proceedings

    3/17

    Presented by Eric Bughin

    May 20 (Thursday) - 10:15 - 11:15

    Presented by Jamil Drarni

    Bas-Relief Ambiguity Reduction in Shape from

    Shadowgrams

    Jamil Drarni1, Sbastien Roy2, Peter Sturm3

    1UNIVERSITDEMONTRAL, CANADA / INRIA-RHNE-ALPES, FRANCE, 2UNIVERSITDEMONTRAL, CANADA, 3INRIA-RHNE-ALPES, FRANCE

    Using shadowgrams for 3D reconstruction is inherently ambiguous. It has be

    shown that this ambiguity has 4 parameters. We show that, when the lightspots are visible to the camera, the number of parameters drops to one.

    3DPVT2010 e-proceedings http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed...

    317 2012-2-17 05:13

    http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceedhttp://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed
  • 8/3/2019 3DPVT2010 e Proceedings

    4/17

    May 20 (Thursday) - 10:15 - 11:15A Probabilistic Approach to ToF and Stereo Data

    FusionCarlo Dal Mutto, Pietro Zanuttigh, Guido Maria Cortelazzo

    UNIVERSITYOFPADOVA, ITALY

    Current 3D video applications require the availability of depth information,

    that can be acquired real-time by stereo vision systems and ToF cameras. Inthis paper, a heterogeneous acquisition system is considered, made of twohigh resolution standard cameras (stereo pair) and one ToF camera. Thestereo system and the ToF camera must be properly calibrated together in

    order to operate jointly. Therefore this work introduces first a generalized

    multi-camera calibration technique which does not exploit only theluminance (color) information, but also the depth information extracted bythe ToF camera. A probabilistic algorithm is then derived in order to obtain

    high quality depth information from the information of both the ToF cameraand the stereo-pair. Experimental results show that the proposed calibrationalgorithm leads to a very accurate calibration suitable for the fusionalgorithm, that allows for precise extraction of the depth information.

    3DPVT2010 e-proceedings http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed...

    417 2012-2-17 05:13

    http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceedhttp://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed
  • 8/3/2019 3DPVT2010 e Proceedings

    5/17

    May 20 (Thursday) - 10:15 - 11:15

    Presented by Toby Collins

    May 20 (Thursday) - 10:15 - 11:15

    Single-view Perspective Shape-from-Texture withFocal Length Estimation: A Piecewise Affine Approach

    Toby Collins1, Jean-Denis Durou2, Pierre Gurdjos3, Adrien Bartoli4

    1UNIVERSITD'AUVERGNE,FRANCE, 2IRIT, UNIVERSITPAUL SABATIER, FRANCE&CMLA, ENS CACHAN, CNRS, UNIVERSUD, FRANCE, 3IRIT, ENSEEIHT, FRANCE,4UNIVERSITD'AUVERGNE, FRANCE

    We present a new formulation to the well known problem of 3D shape-from-texture from a single image, but one which is able to handle

    uncalibrated perspective cameras. Contrary to previous methods, we cast thetask as multi-plane based camera pose estimation whereby the informationprovided by a textured surface makes it possible to perform shape-from-texture and camera focal length estimation jointly. We show that by

    approximating global perspective by local scaled orthography (which holdsoften in practical cases) we can acquire depth, surface orientation and focallength from a single image in closed form. This advances state-of-the-art

    where a calibrated camera is nearly always assumed in order to compute 3Dshape from a single image.

    A statistical illumination-reflectance separation

    algorithm for the multispectral acquisition of frescoes

    Anna Paviotti, Guido Maria Cortelazzo

    3DPVT2010 e-proceedings http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed...

    517 2012-2-17 05:13

    http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceedhttp://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed
  • 8/3/2019 3DPVT2010 e Proceedings

    6/17

    Presented by Guido Maria Cortelazzo

    UNIVERSITYOFPADOVA, ITALY

    The multispectral acquisition of frescoes poses unsolved challenges, the maindifficulty being that it is often impos- sible to measure the reference whitesignal. We propose a statistical method to estimate the illumination directlyfrom the color signal, based on a modification of the RANSAC al- gorithm.

    We apply our method to the estimation of the light- ing field of threepaintings by contemporary Italian artists and of a fresco of the Castello delBuonconsiglio in Trento (Italy), for which a ground truth was available.Quanti- tative results show that the performance of our method is good in

    terms of the relative mean error on illumination and reflectance, while themaximum errors are sometimes significant.

    3DPVT2010 e-proceedings http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed...

    617 2012-2-17 05:13

    http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceedhttp://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed
  • 8/3/2019 3DPVT2010 e Proceedings

    7/17

    May 20 (Thursday) - 10:15 - 11:15

    Presented by Roberto Toldo

    Real-time Incremental J-Linkage for RobustMultiple Structures Estimation

    Roberto Toldo, Andrea Fusiello

    DIPARTIMENTODIINFORMATICA - UNIVERSITDIVERONA, ITALY

    This paper describes an incremental, real-time implementation of J-linkage, aprocedure that can detect multiple instances of a model from data corruptedby noise and outliers. It works in real-time, thanks to several approximationsthat have been introduced to get around the quadratic complexity of the

    original algorithm.

    3D Scanning Methods and Devices

    3DPVT2010 e-proceedings http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed...

    717 2012-2-17 05:13

    http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceedhttp://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed
  • 8/3/2019 3DPVT2010 e Proceedings

    8/17

    May 20 (Thursday) - 10:15 - 11:15

    Presented by David Gallup

    May 20 (Thursday) - 10:15 - 11:15

    A Heightmap Model for Efficient 3DReconstruction from Street-Level Video

    David Gallup1, Jan-Michael Frahm1, Marc Pollefeys2

    1UNIVERSITYOFNORTHCAROLINA, USA, 2ETH ZURICH, SWITZERLAND

    This paper introduces a fast approach for automatic dense large-scale 3Durban reconstruction from video. The presented system uses a novelmulti-view depthmap fusion algorithm where the surface is represented by a

    heightmap. While this model seems to be a more natural fit to aerial andsatellite data, we have found it to also be a powerful representation forground-level reconstructions. It has the advantage of producing purelyvertical facades, and it also yields a continuous surface without holes.

    Compared to more general 3D reconstruction methods, our algorithm ismore efficient, uses less memory, and produces more compact models atthe expense of losing some detail. Our GPU implementation can compute a

    200x200 heightmap from 64 depthmaps in just 92 milliseconds. Wedemonstrate our system on a variety of challenging ground-level datasetsincluding large buildings, residential houses, and store front facadesobtaining clean, complete, compact, and visually pleasing 3D models.

    ?

    Stripe Propagation for Color Encoded Structured

    Light

    3DPVT2010 e-proceedings http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed...

    817 2012-2-17 05:13

    http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceedhttp://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed
  • 8/3/2019 3DPVT2010 e Proceedings

    9/17

    Presented by Michael Burisch

    May 20 (Thursday) - 10:15 - 11:15

    Michael Burisch, David Guerrero Ichaso, Arjan Kuijper

    FRAUNHOFERINSTITUTEFORCOMPUTERGRAPHICSRESEARCHIGD, GERMANY

    We present a 3D scanning system using color encoded structured light,which is able to reconstruct a surface using only a single image. The work is

    focused on exploiting the captured data and reconstructing as many pointsas possible. Therefore a multi-stage method is presented to reconstruct thesurface from the captured pattern. It consists of i) a robust edge detectionstep, ii) a color decoding using feedback from previous stripes and iii) a

    propagation step to detect errors and propagate detected stripes. Usingfeedback from neighboring stripes and propagation along the stripedirection during color detection and sequence decoding, the system is ableto recover areas where regular approaches fail, such as small bridges in

    geometry, and detect and correct false classifications. Depending on surfacecomplexity we were able to achieve a significant increase in the amount ofreconstructable points.

    Active Triangulation in the Outdoors: APhotometric Analysis

    David Ilstrup, Roberto Manduchi

    UC-SANTA CRUZ, USA

    3DPVT2010 e-proceedings http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed...

    917 2012-2-17 05:13

    http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceedhttp://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed
  • 8/3/2019 3DPVT2010 e Proceedings

    10/17

    Presented by David Ilstrup

    May 20 (Thursday) - 10:15 - 11:15

    Active triangulation is a well established technique for collecting rangepoints. This work performs a photometric analysis of relative irradianceexpected at the camera sensor as a result of intended operating conditionsand device specifications including laser power. The limiting effects of eye

    safety compliance, minimum realizable shutter times and pixel bit depth forlinear response cameras are considered. Quantitative results are establisheddetermining dynamic range requirements on the camera, when exposurecontrol is needed, and when laser return can be expected to produce the

    brightest pixels in the image.

    Urban Scene Extraction from Mobile Ground

    Based LiDAR Data

    Joseph Lam1, Michael Greenspan1, Robin Harrap1, Kresimir Kusevic2, Paul Mrstik2

    1QUEEN'SUNIVERSITY, CANADA, 2TERRAPOINTCANADA INC.

    Efficient methods for extracting urban scene structures from 3D data isimportant when dealing with the high volume data collected from mobileterrestrial LiDAR. Rather than searching for primitive shapes directly in theraw 3D data, we demonstrate that the road can be used as a cue for

    effectively localizing urban scene structures. Road extraction is done bydividing the road into many small sections, and Kalman filtering is then used

    3DPVT2010 e-proceedings http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed...

    1017 2012-2-17 05:13

    http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceedhttp://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed
  • 8/3/2019 3DPVT2010 e Proceedings

    11/17

    Presented by Joseph Lam

    May 20 (Thursday) - 10:15 - 11:15

    to track changes of road parameters using a dynamic model. By limiting thesearch space along the extracted road and using dimensional constraints,near-road structures such as posts and power line are easily segmented in anefficient manner. The algorithm performs consistently well on many different

    city scenes, with roads segmented accurately in an efficient manner andposts extracted even in the presence of other objects such as cars and trees.

    3D Shape Retrieval and Recognition

    Ridge Walking for 3D Surface Segmentation

    Andrew Willis, Beibei Zhou

    UNIVERSITYOFNORTHCAROLINAATCHARLOTTE, USA

    This paper describes a new 3D surface segmentation algorithm thatseparates a closed surfaces into regions by computing surface contours thattraverse surface ridge structures. We refer to the approach as ridge-walkingsince these contours tend to follow convex ridge-like structures and/or

    concave valley-like structures present within the geometry of the model.Segmentation is achieved by solving for closed ridge contours on thesurface, each of which serves to divide the surface into two disjoint regions.

    3DPVT2010 e-proceedings http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed...

    1117 2012-2-17 05:13

    http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceedhttp://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed
  • 8/3/2019 3DPVT2010 e Proceedings

    12/17

    Presented by Andrew Willis

    May 20 (Thursday) - 10:15 - 11:15

    Results for three different segmentation approaches based on this approachare compared: (1) concave ridge walking, (2) convex ridge walking and (3)mixed concave/convex ridge walking. We also compare our results withother leading segmentation methods on standard data sets as well as new

    datasets that provide important and interesting new challenges.

    Conformal mapping-based 3D face recognition

    Przemyslaw Szeptycki1, Mohsen Ardabilian1, Liming Chen1, Wei Zeng2, Davig Gu2, Dimitris Samaras2

    1MI DEPARTMENT, LIRIS LABORATORY, ECOLECENTRALEDELYON, FRANCE, 2COMPUTERSCIENCE DEPARTMENT, STONYBROOKUNIVERSITY, USA

    In this paper we present a conformal mapping-based approach for 3D facerecognition. The proposed approach makes use of conformal UVparameterization for mapping purpose and Shape Index decomposition forsimilarity measurement. Indeed, according to conformal geometry theory,

    each 3D surface with disk topology can be mapped onto a 2D domainthrough a global optimization, resulting in a diffeomorphism, i.e.,

    one-to-one and onto. This allows us to reduce the 3D surface matchingproblem to a 2D image matching one by comparing the corresponding 2D

    conformal geometric maps. To deal with facial expressions, the Mobiustransformation of UV conformal space has been used to 'compress' facemimic region. Rasterized images are used as an input for (2D)2PCArecognition algorithm.

    3DPVT2010 e-proceedings http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed...

    1217 2012-2-17 05:13

    http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceedhttp://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed
  • 8/3/2019 3DPVT2010 e Proceedings

    13/17

    Presented by Przemyslaw Szeptycki

    May 20 (Thursday) - 10:15 - 11:15ICP Fusion Techniques for 3D Face Recognition

    Robert McKeon, Patrick Flynn

    UNIVERSITYOFNOTREDAME, USA

    The 3D shape of the face has been shown to be a viable and robust biometric for security

    applications. Many state of the art techniques use Iterative Closest Point (ICP) for 3D facematching. We propose and explore several optimizations of the ICP-based matching techniquerelating to the processing of multiple regions and the fusion of region matching scoresobtained from ICP alignment. The optimizations explored included: (i) the symmetric use of

    probe and gallery face regions as ICP s model and data shapes, enabling score fusion; (ii)gallery and probe region matching score normalization; (iii) region selection based on facedata centroid rather than the nose tip, and (iv) region weighting. As a result of theseoptimizations, the rank-one recognition rate for a canonical matching experiment improved

    from 96.4% to 98.6%, and the True Accept Rate (TAR) at 0.1% False Accept Rate (FAR)

    improved from 90.4% to 98.5%.

    3DPVT2010 e-proceedings http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed...

    1317 2012-2-17 05:13

    http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceedhttp://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed
  • 8/3/2019 3DPVT2010 e Proceedings

    14/17

    Presented by Robert McKeon

    May 20 (Thursday) - 10:15 - 11:15

    Medical Imaging and Augmented Reality

    Indoor Localization Algorithms for a Human-

    Operated Backpack System

    George Chen, John Kua, Stephen Shum, Nikhil Naikal, Matthew Carlberg, Avideh Zakhor

    UC BERKELEYVIDEOANDIMAGEPROCESSINGLAB, USA

    Automated 3D modeling of building interiors is useful in applications such asvirtual reality and entertainment. Using a human-operated backpack system

    equipped with 2D laser scanners and inertial measurement units, we developfour scan-matching-based algorithms to localize the backpack and comparetheir performance and tradeoffs. We present results for two datasets of a

    30-meter-long indoor hallway and compare one of the best performinglocalization algorithms with a visual-odometry-based method. We find thatour scan-matching-based approach results in comparable or higheraccuracy.

    3DPVT2010 e-proceedings http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed...

    1417 2012-2-17 05:13

    http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceedhttp://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed
  • 8/3/2019 3DPVT2010 e Proceedings

    15/17

    Presented by Avideh Zakhor

    May 20 (Thursday) - 10:15 - 11:15Virtu4D: a Real-time Virtualization of Reality

    Mourad Boufarguine1, Malek Baklouti1, Frederic Precioso2, Vincent Guitteny1

    1THALESSECURITY SOLUTIONSANDSERVICES- RESEARCHDEPARTEMENT, FRANCE, 2ETIS ENSEA/CNRS/UNIV. CERGY-PONTOISE, FRANCE

    In video surveillance systems, when dealing with dynamic complex scenes,

    processing the information coming from multiple cameras and fusing theminto a comprehensible environment is a challenging task. This work

    addresses the issue of providing a global and reliable representation of themonitored environment aiming at enhancing the perception and minimizingthe operator's effort. The proposed system Virtu4D is based on 3Dcomputer vision and virtual reality techniques and takes benefit from both

    the "real" and the "virtual" worlds offering a unique perception of the scene.This paper presents a short overview of the framework along with thedifferent components of the design space: Video Model Layout, VideoProcessing and Immersive Model Generation. The final interface gathers the

    2D information in the 3D context but also offers a complete 3D

    representation of the dynamic environment allowing a free intuitive 3Dnavigation.

    3DPVT2010 e-proceedings http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed...

    1517 2012-2-17 05:13

    http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceedhttp://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed
  • 8/3/2019 3DPVT2010 e Proceedings

    16/17

    Presented by Mourad Boufarguine

    May 20 (Thursday) - 10:15 - 11:15Sparse Parallel Electronic Bowel Cleansing in CT

    Colonography

    Richard Boyes, Xujiong Ye, Gareth Beddoe, Greg Slabaugh

    MEDICSIGHTPLC, UNITEDKINGDOM

    We present a technique for storing the sparse data that often occurs whenprocessing three dimensional medical images. The technique uses raster scan

    order to store the one dimensional volume indexes of each pixel location,and stores an inverted copy of these indexes for fast lookup. The invertedindex is stored as a Judy array which is shown to be highly efficient in lookuptimes while using very little memory compared to hash tables. We

    demonstrate the efficiency of the data structure by performing partialvolume segmentation and digital removal of oral contrast agent within CTColonography (CTC).

    3DPVT2010 e-proceedings http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed...

    1617 2012-2-17 05:13

    http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceedhttp://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed
  • 8/3/2019 3DPVT2010 e Proceedings

    17/17

    Presented by Xujiong Ye

    3DPVT2010 e-proceedings http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed...

    1717 2012-2-17 05:13

    http://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceedhttp://campwww.informatik.tu-muenchen.de/3DPVT2010/data/media/e-proceed