Solving Vision Tasks with variational methods on the GPU

Size: px
Start display at page:

Download "Solving Vision Tasks with variational methods on the GPU"

Transcription

1 Solving Vision Tasks with variational methods on the GPU Horst Bischof Inst. f. Computer Graphics and Vision Graz University of Technology Joint work with Thomas Pock, Markus Unger, Arnold Irschara and the ICG team 1

2 GPU-Speed 2

3 Recent GPU-Hardware Executes hundreds of parallel threads 32/64 bit floating point arithmetic, lots of main memory Rich set of general purpose arithmetic operations Features accessible via C-like languages (Cg, CUDA) (Supercomputer in pocket format) Nvidia GTX 285 Nvidia Tesla S1070 3

4 Embedded Systems Drivers: GPUs on mobile Phones Mobile Games Nvidia Tegra 4

5 What type of algorithms? Highly parallel (parallelizable) Good mathematical properties Measure of quality Variety of problems Variational Methods 5

6 Outline Total Variation Basics Applications Segmentation Optical Flow Stereo & 3D Conclusion 6

7 Computer Vision is Ill-Posed [1] Restrict the space of possible solutions by an a-priori asumption of the solution The Bayesian framework is often used to estimate the unknown quantities Equivalent to the Variational approach [1] J. Hadamard. Sur les problémes aux dérivées partielles et leur signification physique

8 Introduction 8

9 What is the TV norm? The TV norm is the L 1 norm of the L 2 vector norm of the image gradient History of L 1 estimation techniques Galileo Galilei, 1632 Laplace, 1793 Huber,

10 Why does it preserve discontinuities? W u dx = 1.0 W u 2 dx = Total Variation has no bias against discontinuities 10

11 Denoising Model of Rudin Osher and Fatemi Defined as the Variational Problem [3],[4] min u W u dx + l 2 W ( u - f ) 2 dx [3] L. Rudin and S. Osher and E. Fatemi. Nonlinear Total Variation Based Noise Removal Algorithms, 1992 [4] A. Chambolle and P. Lions, Image Recovery via Total Variation Minimization and Applications,

12 First Convex Example Total Variation 12

13 Numerical Methods Euler-Lagrange equations of the ROF model E E u l 2 = u dw + ( u - u0 ) dw 2 W = - Ł u u W + l ł ( u - u ) 0 0 = (Primal Formulation) Problem: EL equation is degenerated in flat regions Replace u by u 2 = u + e 2 e But decreases ability to preserve sharp edges 13

14 Variational Denoising Chambolle s projection algorithm Duality based formulation of the unconstrained formulation u max r p 1 r p u max u r p dw - r 0 p 1 2 W 1 a W r 2 ( p) dw Dual Formulation is quadratic in p, but has a bad constraint. Key observation of A. Chambolle: The Lagrange multipliers for the constrained Euler-Lagrange equation can be eliminated. Simple fixed point iteration scheme. r n 1 r n p + t p - u0 r n+ 1 p = Ła ł 1 r n 1+ t p - u0 Ła ł A. Chambolle. An Algorithm for Total Variation Minimization and Applications. J. Math. Imaging,

15 Primal ROF Primal versus Dual Non-smooth optimization problem Hard to optimize Dual ROF Smooth optimization problem with constraints Easy to optimize What about Primal-Dual? Make use of both, primal and dual More general than pure primal or dual Extremely fast primal-dual algorithm [Pock, Cremers, Bischof, Chambolle, 2009] 15

16 Outline Total Variation Basics Applications Segmentation Optical Flow Stereo & 3D Conclusion 16

17 Variational Image Denoising TV-L2 Denoising u Reconstructed Image f Original images with artifacts λ...regularization parameter Ω Image Domain TV-L1 Denoising 17

18 Variational Image Denoising TV-L2 18

19 Variational Image Denoising TV-L1 19

20 Link from Denoising to Segmentation 20

21 ( ) Proposed Energy Extend weighted TV with spatially varying data term: min u W ( x) u dx + l( x) u - f is provided by the user: f = l( x) = 0... Information not used 0 < l( x) <... Weak constraints l( x) fi... Hard constraints g x can also be modified by the user Functional remains convex Convex formulation of GAC g 0 background 1 foreground M. Unger, T. Pock, W. Trobin, D. Cremers, and H.Bischof. TV-Seg - interactive total variation based image segmentation. BMVC W f dx

22 Evolution of u using hard constraints iterations 22

23 With Color Model 23

24 Brain Segmentation 24

25 25

26 3D Interactive Segmentation 26

27 Texture Features Interactive texture segmentation using HOG features and on-line random forests Features pre-calculated Texture descriptor learned on-line using RF J. Santner, M. Unger, T. Pock, Ch. Leistner, A. Saffari, and H. Bischof. Interactive texture segmentation using random forests and total variation. In BMVC'09,

28 Results A. Saffari, C. Leistner, J. Santner, M. Godec, and H. Bischof. On-line random forests. In 3rd IEEE On-line Learning for Computer Vision Workshop. IEEE,

29 Outline Total Variation Basics Applications Segmentation Optical Flow Stereo & 3D Conclusion 29

30 TV-L 1 Optical Flow We use a robust variant of the Horn-Schunck formulation [8] Total Variation Regularization and L 1 data term Total Variation of Flow Sophisticated optimization techniques are needed! We have developed fast primal-dual schemes Implemented on the GPU L 1 norm of Optical Flow Constraint [8] B.K. Horn and B.G. Schunck. Determinig Optical Flow. Artificial Intelligence,

31 Huber Norm + Anisotropic Replace stair-casing afflicted isotropic Total Variation (TV) by a robust penalty function initially proposed by Huber. Incorporate directional information yielding an anisotropic Huber regularity. Huber Anisotropic M. Werlberger, W. Trobin, T. Pock, A. Wedel, D. Cremers, and H. Bischof. Anisotropic Huber-L1 optical flow. In BMVC'09,

32 Performance Evaluation TV-L 1 Optical Flow Implemented in CUDA 3.0 Computed on nvidia GeForce GTX 280 GPU 50 iterations / pyramid level Image Size Frames per Second 128x x x

33 Evaluation of optical flow methods Input Images Ground Truth Challenging Benchmark Dataset: 33

34 Results 34

35 Optical Flow Benchmark 35

36 Optical Flow for Driver Assistence (A. Wedel, Daimler) 36

37 Tracking using only Flow 37

38 Tracking as 3D Volume Segmentation Use Segmentation in 3D Minimize Volume M. Professor Unger, Horst Bischof T. Mauthner, Horst Cerjak, T. Pock, and H. Bischof. Tracking as segmentation of spatial-temporal volumes Variational by anisotropic Methods weighted & GPU TV. In EMCVPR 2009 pp Springer,

39 Results 39

40 Outline Total Variation Basics Applications Segmentation Optical Flow Stereo & 3D Conclusion 40

41 Total Variation energy functional Total Variation regularization Data term Data term potentially non-convex Defines domain of application Denoising Stereo T. Pock, T. Schönemann, G. Graber, H. Bischof, and D. Cremers. A convex formulation of continuous multi-label problems. ECCV08 41

42 Finding a Convex Representation Original non-convex problem New convex formulation Theorem: Minimizing is equivalent to minimizing 42

43 Convex Relaxation Theory due to Alberti and Bouchitte (2002) Consider the function u(x) as a surface in higher dimensions (functional lifting) Consider the maximal flux of a vector field going through the surface T. Pock, D. Cremers, H. Bischof, and A. Chambolle. An algorithm for minimizing the Mumford-Shah functional. ICCV,

44 Stereo results We applied our method to stereo problems Data term: absolute differences 44

45 Qualitative comparison to Ishikawa Tsukuba data Ground truth 45

46 Quantitative comparison to Ishikawa Tsukuba data set, (size=384x288) Results comparable to 16 connected graph The proposed algorithm is 20 times faster Requires only 3,6% of the memory Can be applied to much larger problems 46

47 3D Reconstruction System Feature Extraction Feature Matching Geometric Verification Track Generation Initial Structure from Motion Bundle Adjustment Dense Matching Range Image Fusion 47

48 Aerial Triangulation (SfM approach) 3.7 Mio. 3D points (SIFT keys) ~4200 measurements / image 48

49 DEPTH map 49

50 DEPTH map 50

51 DEPTH map 51

52 Jakomini Platz 52

53 Depth Map Fusion Task: Compute Digital Surface Model (DSM) from range images High overlap in aerial images One position is seen from 8-15 images Exploit redundancy to improve the DSM Robust fusion of range images to a single DSM using a Total Generalized Variation functional 53

54 Results 54

55 Fusion of real Data 55

56 Conclusion Total Variation is powerful Convex Optimization Variety of Applications Fast using GPUs General Framework, easy to extend Couple different modules 56

57 Videos/Code/Papers see 57

58 Acknowledgments Funding provided by: Austrian Joint Research Project Cognitive Vision under sub-projects S9103-N03 and S9104-N04 Doctoral Program Confluence of Vision and Graphics funded by Austrian Science Found FIT-IT Program funded by BMVIT under Project VMGPU Ludwig Boltzmann Inst. on Forensic Radiology 58

3D Computer Vision. Dense 3D Reconstruction II. Prof. Didier Stricker. Christiano Gava

3D Computer Vision. Dense 3D Reconstruction II. Prof. Didier Stricker. Christiano Gava 3D Computer Vision Dense 3D Reconstruction II Prof. Didier Stricker Christiano Gava Kaiserlautern University http://ags.cs.uni-kl.de/ DFKI Deutsches Forschungszentrum für Künstliche Intelligenz http://av.dfki.de

More information

Motion Estimation with Adaptive Regularization and Neighborhood Dependent Constraint

Motion Estimation with Adaptive Regularization and Neighborhood Dependent Constraint 0 Digital Image Computing: Techniques and Applications Motion Estimation with Adaptive Regularization and Neighborhood Dependent Constraint Muhammad Wasim Nawaz, Abdesselam Bouzerdoum, Son Lam Phung ICT

More information

Notes 9: Optical Flow

Notes 9: Optical Flow Course 049064: Variational Methods in Image Processing Notes 9: Optical Flow Guy Gilboa 1 Basic Model 1.1 Background Optical flow is a fundamental problem in computer vision. The general goal is to find

More information

On a first-order primal-dual algorithm

On a first-order primal-dual algorithm On a first-order primal-dual algorithm Thomas Pock 1 and Antonin Chambolle 2 1 Institute for Computer Graphics and Vision, Graz University of Technology, 8010 Graz, Austria 2 Centre de Mathématiques Appliquées,

More information

Interactive Texture Segmentation using Random Forests and Total Variation

Interactive Texture Segmentation using Random Forests and Total Variation SANTNER ET AL.: INTERACTIVE TEXTURE SEGMENTATION 1 Interactive Texture Segmentation using Random Forests and Total Variation Jakob Santner santner@icg.tugraz.at Markus Unger unger@icg.tugraz.at Thomas

More information

A Duality Based Algorithm for TV-L 1 -Optical-Flow Image Registration

A Duality Based Algorithm for TV-L 1 -Optical-Flow Image Registration A Duality Based Algorithm for TV-L 1 -Optical-Flow Image Registration Thomas Pock 1, Martin Urschler 1, Christopher Zach 2, Reinhard Beichel 3, and Horst Bischof 1 1 Institute for Computer Graphics & Vision,

More information

Video Super Resolution using Duality Based TV-L 1 Optical Flow

Video Super Resolution using Duality Based TV-L 1 Optical Flow Video Super Resolution using Duality Based TV-L 1 Optical Flow Dennis Mitzel 1,2, Thomas Pock 3, Thomas Schoenemann 1 Daniel Cremers 1 1 Department of Computer Science University of Bonn, Germany 2 UMIC

More information

Project Updates Short lecture Volumetric Modeling +2 papers

Project Updates Short lecture Volumetric Modeling +2 papers Volumetric Modeling Schedule (tentative) Feb 20 Feb 27 Mar 5 Introduction Lecture: Geometry, Camera Model, Calibration Lecture: Features, Tracking/Matching Mar 12 Mar 19 Mar 26 Apr 2 Apr 9 Apr 16 Apr 23

More information

Vertebrae Segmentation in 3D CT Images based on a Variational Framework

Vertebrae Segmentation in 3D CT Images based on a Variational Framework Vertebrae Segmentation in 3D CT Images based on a Variational Framework Kerstin Hammernik, Thomas Ebner, Darko Stern, Martin Urschler, and Thomas Pock Abstract Automatic segmentation of 3D vertebrae is

More information

Part II: Modeling Aspects

Part II: Modeling Aspects Yosemite test sequence Illumination changes Motion discontinuities Variational Optical Flow Estimation Part II: Modeling Aspects Discontinuity Di ti it preserving i smoothness th tterms Robust data terms

More information

Supplementary Material Estimating Correspondences of Deformable Objects In-the-wild

Supplementary Material Estimating Correspondences of Deformable Objects In-the-wild Supplementary Material Estimating Correspondences of Deformable Objects In-the-wild Yuxiang Zhou Epameinondas Antonakos Joan Alabort-i-Medina Anastasios Roussos Stefanos Zafeiriou, Department of Computing,

More information

ILLUMINATION ROBUST OPTICAL FLOW ESTIMATION BY ILLUMINATION-CHROMATICITY DECOUPLING. Sungheon Park and Nojun Kwak

ILLUMINATION ROBUST OPTICAL FLOW ESTIMATION BY ILLUMINATION-CHROMATICITY DECOUPLING. Sungheon Park and Nojun Kwak ILLUMINATION ROBUST OPTICAL FLOW ESTIMATION BY ILLUMINATION-CHROMATICITY DECOUPLING Sungheon Park and Nojun Kwak Graduate School of Convergence Science and Technology, Seoul National University, Korea

More information

Variational Methods II

Variational Methods II Mathematical Foundations of Computer Graphics and Vision Variational Methods II Luca Ballan Institute of Visual Computing Last Lecture If we have a topological vector space with an inner product and functionals

More information

Outline. Level Set Methods. For Inverse Obstacle Problems 4. Introduction. Introduction. Martin Burger

Outline. Level Set Methods. For Inverse Obstacle Problems 4. Introduction. Introduction. Martin Burger For Inverse Obstacle Problems Martin Burger Outline Introduction Optimal Geometries Inverse Obstacle Problems & Shape Optimization Sensitivity Analysis based on Gradient Flows Numerical Methods University

More information

Chapter 2 Optical Flow Estimation

Chapter 2 Optical Flow Estimation Chapter 2 Optical Flow Estimation Abstract In this chapter we review the estimation of the two-dimensional apparent motion field of two consecutive images in an image sequence. This apparent motion field

More information

A Duality Based Algorithm for TV-L 1 -Optical-Flow Image Registration

A Duality Based Algorithm for TV-L 1 -Optical-Flow Image Registration A Duality Based Algorithm for TV-L 1 -Optical-Flow Image Registration Thomas Pock 1,MartinUrschler 1, Christopher Zach 2, Reinhard Beichel 3, and Horst Bischof 1 1 Institute for Computer Graphics & Vision,

More information

Geometric and Semantic 3D Reconstruction: Part 4A: Volumetric Semantic 3D Reconstruction. CVPR 2017 Tutorial Christian Häne UC Berkeley

Geometric and Semantic 3D Reconstruction: Part 4A: Volumetric Semantic 3D Reconstruction. CVPR 2017 Tutorial Christian Häne UC Berkeley Geometric and Semantic 3D Reconstruction: Part 4A: Volumetric Semantic 3D Reconstruction CVPR 2017 Tutorial Christian Häne UC Berkeley Dense Multi-View Reconstruction Goal: 3D Model from Images (Depth

More information

Exploiting Redundancy for Aerial Image Fusion using Convex Optimization

Exploiting Redundancy for Aerial Image Fusion using Convex Optimization Exploiting Redundancy for Aerial Image Fusion using Convex Optimization Stefan Kluckner, Thomas Pock and Horst Bischof Institute for Computer Graphics and Vision Graz University of Technology, Austria

More information

Continuous and Discrete Optimization Methods in Computer Vision. Daniel Cremers Department of Computer Science University of Bonn

Continuous and Discrete Optimization Methods in Computer Vision. Daniel Cremers Department of Computer Science University of Bonn Continuous and Discrete Optimization Methods in Computer Vision Daniel Cremers Department of Computer Science University of Bonn Oxford, August 16 2007 Segmentation by Energy Minimization Given an image,

More information

Technion - Computer Science Department - Tehnical Report CIS

Technion - Computer Science Department - Tehnical Report CIS Over-Parameterized Variational Optical Flow Tal Nir Alfred M. Bruckstein Ron Kimmel {taln, freddy, ron}@cs.technion.ac.il Department of Computer Science Technion Israel Institute of Technology Technion

More information

Robust Trajectory-Space TV-L1 Optical Flow for Non-rigid Sequences

Robust Trajectory-Space TV-L1 Optical Flow for Non-rigid Sequences Robust Trajectory-Space TV-L1 Optical Flow for Non-rigid Sequences Ravi Garg, Anastasios Roussos, and Lourdes Agapito Queen Mary University of London, Mile End Road, London E1 4NS, UK Abstract. This paper

More information

Optical Flow and Dense Correspondence. Daniel Cremers Computer Science & Mathematics TU Munich

Optical Flow and Dense Correspondence. Daniel Cremers Computer Science & Mathematics TU Munich Optical Flow and Dense Correspondence Computer Science & Mathematics TU Munich Correspondence Problems in Vision 2 Shape Similarity & Shape Matching 3 Overview Optical Flow Estimation Dense Elastic Shape

More information

TVSeg - Interactive Total Variation Based Image Segmentation

TVSeg - Interactive Total Variation Based Image Segmentation TVSeg - Interactive Total Variation Based Image Segmentation Markus Unger 1, Thomas Pock 1,2, Werner Trobin 1, Daniel Cremers 2, Horst Bischof 1 1 Inst. for Computer Graphics & Vision, Graz University

More information

PROST: Parallel Robust Online Simple Tracking

PROST: Parallel Robust Online Simple Tracking PROST: Parallel Robust Online Simple Tracking Jakob Santner Christian Leistner Amir Saffari Thomas Pock Horst Bischof Institute for Computer Graphics and Vision, Graz University of Technology {santner,leistner,saffari,pock,bischof}@icg.tugraz.at

More information

Intrinsic3D: High-Quality 3D Reconstruction by Joint Appearance and Geometry Optimization with Spatially-Varying Lighting

Intrinsic3D: High-Quality 3D Reconstruction by Joint Appearance and Geometry Optimization with Spatially-Varying Lighting Intrinsic3D: High-Quality 3D Reconstruction by Joint Appearance and Geometry Optimization with Spatially-Varying Lighting R. Maier 1,2, K. Kim 1, D. Cremers 2, J. Kautz 1, M. Nießner 2,3 Fusion Ours 1

More information

TV-L1 Optical Flow Estimation

TV-L1 Optical Flow Estimation 214/7/1 v.5 IPOL article class Published in Image Processing On Line on 213 7 19. Submitted on 212 7 5, accepted on 212 1 29. ISSN 215 1232 c 213 IPOL & the authors CC BY NC SA This article is available

More information

An Evaluation of Robust Cost Functions for RGB Direct Mapping

An Evaluation of Robust Cost Functions for RGB Direct Mapping An Evaluation of Robust Cost Functions for RGB Direct Mapping Alejo Concha and Javier Civera Abstract The so-called direct SLAM methods have shown an impressive performance in estimating a dense 3D reconstruction

More information

Numerical Methods on the Image Processing Problems

Numerical Methods on the Image Processing Problems Numerical Methods on the Image Processing Problems Department of Mathematics and Statistics Mississippi State University December 13, 2006 Objective Develop efficient PDE (partial differential equations)

More information

A MULTI-RESOLUTION APPROACH TO DEPTH FIELD ESTIMATION IN DENSE IMAGE ARRAYS F. Battisti, M. Brizzi, M. Carli, A. Neri

A MULTI-RESOLUTION APPROACH TO DEPTH FIELD ESTIMATION IN DENSE IMAGE ARRAYS F. Battisti, M. Brizzi, M. Carli, A. Neri A MULTI-RESOLUTION APPROACH TO DEPTH FIELD ESTIMATION IN DENSE IMAGE ARRAYS F. Battisti, M. Brizzi, M. Carli, A. Neri Università degli Studi Roma TRE, Roma, Italy 2 nd Workshop on Light Fields for Computer

More information

Improving Motion Estimation Using Image-Driven Functions and Hybrid Scheme

Improving Motion Estimation Using Image-Driven Functions and Hybrid Scheme Improving Motion Estimation Using Image-Driven Functions and Hybrid Scheme Duc Dung Nguyen and Jae Wook Jeon Department of Electrical and Computer Engineering, Sungkyunkwan University, Korea nddunga3@skku.edu,

More information

Tracking and Structure from Motion

Tracking and Structure from Motion Master s Thesis Winter Semester 2009/2010 Tracking and Structure from Motion Author: Andreas Weishaupt Supervisor: Prof. Pierre Vandergheynst Assistants: Luigi Bagnato Emmanuel D Angelo Signal Processing

More information

Reconstructing Reflective and Transparent Surfaces from Epipolar Plane Images

Reconstructing Reflective and Transparent Surfaces from Epipolar Plane Images Sven Wanner and Bastian Goldlücke Heidelberg Collaboratory for Image Processing 1. Light Fields and Epipolar Plane Images 1. Light Fields and Epipolar Plane Images Light Fields as a dense sampling of a

More information

Lecture 10 Multi-view Stereo (3D Dense Reconstruction) Davide Scaramuzza

Lecture 10 Multi-view Stereo (3D Dense Reconstruction) Davide Scaramuzza Lecture 10 Multi-view Stereo (3D Dense Reconstruction) Davide Scaramuzza REMODE: Probabilistic, Monocular Dense Reconstruction in Real Time, ICRA 14, by Pizzoli, Forster, Scaramuzza [M. Pizzoli, C. Forster,

More information

4/13/ Introduction. 1. Introduction. 2. Formulation. 2. Formulation. 2. Formulation

4/13/ Introduction. 1. Introduction. 2. Formulation. 2. Formulation. 2. Formulation 1. Introduction Motivation: Beijing Jiaotong University 1 Lotus Hill Research Institute University of California, Los Angeles 3 CO 3 for Ultra-fast and Accurate Interactive Image Segmentation This paper

More information

Contents. I Basics 1. Copyright by SIAM. Unauthorized reproduction of this article is prohibited.

Contents. I Basics 1. Copyright by SIAM. Unauthorized reproduction of this article is prohibited. page v Preface xiii I Basics 1 1 Optimization Models 3 1.1 Introduction... 3 1.2 Optimization: An Informal Introduction... 4 1.3 Linear Equations... 7 1.4 Linear Optimization... 10 Exercises... 12 1.5

More information

LECTURE 13: SOLUTION METHODS FOR CONSTRAINED OPTIMIZATION. 1. Primal approach 2. Penalty and barrier methods 3. Dual approach 4. Primal-dual approach

LECTURE 13: SOLUTION METHODS FOR CONSTRAINED OPTIMIZATION. 1. Primal approach 2. Penalty and barrier methods 3. Dual approach 4. Primal-dual approach LECTURE 13: SOLUTION METHODS FOR CONSTRAINED OPTIMIZATION 1. Primal approach 2. Penalty and barrier methods 3. Dual approach 4. Primal-dual approach Basic approaches I. Primal Approach - Feasible Direction

More information

Real-Time Simultaneous 3D Reconstruction and Optical Flow Estimation

Real-Time Simultaneous 3D Reconstruction and Optical Flow Estimation Real-Time Simultaneous 3D Reconstruction and Optical Flow Estimation Menandro Roxas Takeshi Oishi Institute of Industrial Science, The University of Tokyo roxas, oishi @cvl.iis.u-tokyo.ac.jp Abstract We

More information

FLaME: Fast Lightweight Mesh Estimation using Variational Smoothing on Delaunay Graphs

FLaME: Fast Lightweight Mesh Estimation using Variational Smoothing on Delaunay Graphs FLaME: Fast Lightweight Mesh Estimation using Variational Smoothing on Delaunay Graphs W. Nicholas Greene Robust Robotics Group, MIT CSAIL LPM Workshop IROS 2017 September 28, 2017 with Nicholas Roy 1

More information

Lecture 10 Dense 3D Reconstruction

Lecture 10 Dense 3D Reconstruction Institute of Informatics Institute of Neuroinformatics Lecture 10 Dense 3D Reconstruction Davide Scaramuzza 1 REMODE: Probabilistic, Monocular Dense Reconstruction in Real Time M. Pizzoli, C. Forster,

More information

Yunyun Yang, Chunming Li, Chiu-Yen Kao and Stanley Osher. Speaker: Chiu-Yen Kao (Math Department, The Ohio State University) BIRS, Banff, Canada

Yunyun Yang, Chunming Li, Chiu-Yen Kao and Stanley Osher. Speaker: Chiu-Yen Kao (Math Department, The Ohio State University) BIRS, Banff, Canada Yunyun Yang, Chunming Li, Chiu-Yen Kao and Stanley Osher Speaker: Chiu-Yen Kao (Math Department, The Ohio State University) BIRS, Banff, Canada Outline Review of Region-based Active Contour Models Mumford

More information

Structured Light II. Thanks to Ronen Gvili, Szymon Rusinkiewicz and Maks Ovsjanikov

Structured Light II. Thanks to Ronen Gvili, Szymon Rusinkiewicz and Maks Ovsjanikov Structured Light II Johannes Köhler Johannes.koehler@dfki.de Thanks to Ronen Gvili, Szymon Rusinkiewicz and Maks Ovsjanikov Introduction Previous lecture: Structured Light I Active Scanning Camera/emitter

More information

COMPUTER VISION > OPTICAL FLOW UTRECHT UNIVERSITY RONALD POPPE

COMPUTER VISION > OPTICAL FLOW UTRECHT UNIVERSITY RONALD POPPE COMPUTER VISION 2017-2018 > OPTICAL FLOW UTRECHT UNIVERSITY RONALD POPPE OUTLINE Optical flow Lucas-Kanade Horn-Schunck Applications of optical flow Optical flow tracking Histograms of oriented flow Assignment

More information

Variational Optical Flow from Alternate Exposure Images

Variational Optical Flow from Alternate Exposure Images Variational Optical Flow from Alternate Exposure Images A. Sellent, M. Eisemann, B. Goldlücke, T. Pock, D. Cremers, M. Magnor TU Braunschweig, University Bonn, TU Graz Two Image Optical Flow Pinpoint sharp

More information

Mathematics in Image Processing

Mathematics in Image Processing Mathematics in Image Processing Michal Šorel Department of Image Processing Institute of Information Theory and Automation (ÚTIA) Academy of Sciences of the Czech Republic http://zoi.utia.cas.cz/ Mathematics

More information

Dense 3D Modelling and Monocular Reconstruction of Deformable Objects

Dense 3D Modelling and Monocular Reconstruction of Deformable Objects Dense 3D Modelling and Monocular Reconstruction of Deformable Objects Anastasios (Tassos) Roussos Lecturer in Computer Science, University of Exeter Research Associate, Imperial College London Overview

More information

A Patch Prior for Dense 3D Reconstruction in Man-Made Environments

A Patch Prior for Dense 3D Reconstruction in Man-Made Environments A Patch Prior for Dense 3D Reconstruction in Man-Made Environments Christian Häne 1, Christopher Zach 2, Bernhard Zeisl 1, Marc Pollefeys 1 1 ETH Zürich 2 MSR Cambridge October 14, 2012 A Patch Prior for

More information

Global Minimization of the Active Contour Model with TV-Inpainting and Two-Phase Denoising

Global Minimization of the Active Contour Model with TV-Inpainting and Two-Phase Denoising Global Minimization of the Active Contour Model with TV-Inpainting and Two-Phase Denoising Shingyu Leung and Stanley Osher Department of Mathematics, UCLA, Los Angeles, CA 90095, USA {syleung, sjo}@math.ucla.edu

More information

Problèmes Inverses et Imagerie. Journées de rentrée des Masters, 6-8 Septembre 2017, IHES, Bures-sur-Yvette France

Problèmes Inverses et Imagerie. Journées de rentrée des Masters, 6-8 Septembre 2017, IHES, Bures-sur-Yvette France Problèmes Inverses et Imagerie Ce que les images ne nous disent pas. Luca Calatroni Centre de Mathematiqués Appliquées (CMAP), École Polytechnique CNRS Journées de rentrée des Masters, 6-8 Septembre 2017,

More information

RegionCut - Interactive Multi-Label Segmentation Utilizing Cellular Automaton

RegionCut - Interactive Multi-Label Segmentation Utilizing Cellular Automaton RegionCut - Interactive Multi-Label Segmentation Utilizing Cellular Automaton Oliver Jakob Arndt, Björn Scheuermann, Bodo Rosenhahn Institut für Informationsverarbeitung (TNT), Leibniz Universität Hannover,

More information

Optimizing Parametric Total Variation Models

Optimizing Parametric Total Variation Models Optimizing Parametric Total Variation Models Strandmark, Petter; Kahl, Fredrik; Overgaard, Niels Christian Published in: [Host publication title missing] DOI:.9/ICCV.9.5459464 9 Link to publication Citation

More information

Motion Cooperation: Smooth Piece-Wise Rigid Scene Flow from RGB-D Images

Motion Cooperation: Smooth Piece-Wise Rigid Scene Flow from RGB-D Images Motion Cooperation: Smooth Piece-Wise Rigid Scene Flow from RGB-D Images Mariano Jaimez1,2 Mohamed Souiai1 Jo rg Stu ckler1 Javier Gonzalez-Jimenez2 Daniel Cremers1 1 Technische Universita t Mu nchen,

More information

3D Computer Vision. Structured Light II. Prof. Didier Stricker. Kaiserlautern University.

3D Computer Vision. Structured Light II. Prof. Didier Stricker. Kaiserlautern University. 3D Computer Vision Structured Light II Prof. Didier Stricker Kaiserlautern University http://ags.cs.uni-kl.de/ DFKI Deutsches Forschungszentrum für Künstliche Intelligenz http://av.dfki.de 1 Introduction

More information

Sublabel Accurate Relaxation of Nonconvex Energies

Sublabel Accurate Relaxation of Nonconvex Energies Sublabel Accurate Relaxation of Nonconvex Energies Thomas Möllenhoff TU München moellenh@in.tum.de Emanuel Laude TU München laudee@in.tum.de Jan Lellmann University of Lübeck lellmann@mic.uni-luebeck.de

More information

Discrete Optimization of Ray Potentials for Semantic 3D Reconstruction

Discrete Optimization of Ray Potentials for Semantic 3D Reconstruction Discrete Optimization of Ray Potentials for Semantic 3D Reconstruction Marc Pollefeys Joined work with Nikolay Savinov, Christian Haene, Lubor Ladicky 2 Comparison to Volumetric Fusion Higher-order ray

More information

Advanced phase retrieval: maximum likelihood technique with sparse regularization of phase and amplitude

Advanced phase retrieval: maximum likelihood technique with sparse regularization of phase and amplitude Advanced phase retrieval: maximum likelihood technique with sparse regularization of phase and amplitude A. Migukin *, V. atkovnik and J. Astola Department of Signal Processing, Tampere University of Technology,

More information

Optical Flow Estimation with CUDA. Mikhail Smirnov

Optical Flow Estimation with CUDA. Mikhail Smirnov Optical Flow Estimation with CUDA Mikhail Smirnov msmirnov@nvidia.com Document Change History Version Date Responsible Reason for Change Mikhail Smirnov Initial release Abstract Optical flow is the apparent

More information

Multiframe Scene Flow with Piecewise Rigid Motion. Vladislav Golyanik,, Kihwan Kim, Robert Maier, Mathias Nießner, Didier Stricker and Jan Kautz

Multiframe Scene Flow with Piecewise Rigid Motion. Vladislav Golyanik,, Kihwan Kim, Robert Maier, Mathias Nießner, Didier Stricker and Jan Kautz Multiframe Scene Flow with Piecewise Rigid Motion Vladislav Golyanik,, Kihwan Kim, Robert Maier, Mathias Nießner, Didier Stricker and Jan Kautz Scene Flow. 2 Scene Flow. 3 Scene Flow. Scene Flow Estimation:

More information

Occlusion-Aware Video Registration for Highly Non-Rigid Objects Supplementary Material

Occlusion-Aware Video Registration for Highly Non-Rigid Objects Supplementary Material Occlusion-Aware Video Registration for Highly Non-Rigid Objects Supplementary Material Bertram Taetz 1,, Gabriele Bleser 1,, Vladislav Golyanik 1, and Didier Stricker 1, 1 German Research Center for Artificial

More information

CS 565 Computer Vision. Nazar Khan PUCIT Lectures 15 and 16: Optic Flow

CS 565 Computer Vision. Nazar Khan PUCIT Lectures 15 and 16: Optic Flow CS 565 Computer Vision Nazar Khan PUCIT Lectures 15 and 16: Optic Flow Introduction Basic Problem given: image sequence f(x, y, z), where (x, y) specifies the location and z denotes time wanted: displacement

More information

Stereo Scene Flow for 3D Motion Analysis

Stereo Scene Flow for 3D Motion Analysis Stereo Scene Flow for 3D Motion Analysis Andreas Wedel Daniel Cremers Stereo Scene Flow for 3D Motion Analysis Dr. Andreas Wedel Group Research Daimler AG HPC 050 G023 Sindelfingen 71059 Germany andreas.wedel@daimler.com

More information

Total Variation Denoising with Overlapping Group Sparsity

Total Variation Denoising with Overlapping Group Sparsity 1 Total Variation Denoising with Overlapping Group Sparsity Ivan W. Selesnick and Po-Yu Chen Polytechnic Institute of New York University Brooklyn, New York selesi@poly.edu 2 Abstract This paper describes

More information

Scale Invariant Optical Flow

Scale Invariant Optical Flow Scale Invariant Optical Flow Li Xu Zhenlong Dai Jiaya Jia Department of Computer Science and Engineering The Chinese University of Hong Kong {xuli,zldai,leojia}@cse.cuhk.edu.hk Abstract. Scale variation

More information

Convex Optimization and Machine Learning

Convex Optimization and Machine Learning Convex Optimization and Machine Learning Mengliu Zhao Machine Learning Reading Group School of Computing Science Simon Fraser University March 12, 2014 Mengliu Zhao SFU-MLRG March 12, 2014 1 / 25 Introduction

More information

A Patch Prior for Dense 3D Reconstruction in Man-Made Environments

A Patch Prior for Dense 3D Reconstruction in Man-Made Environments A Patch Prior for Dense 3D Reconstruction in Man-Made Environments Christian Häne 1 Christopher Zach 2 Bernhard Zeisl 1 Marc Pollefeys 1 ETH Zürich 1 Switzerland {chaene, zeislb, pomarc}@inf.ethz.ch Microsoft

More information

Camera Drones Lecture 3 3D data generation

Camera Drones Lecture 3 3D data generation Camera Drones Lecture 3 3D data generation Ass.Prof. Friedrich Fraundorfer WS 2017 Outline SfM introduction SfM concept Feature matching Camera pose estimation Bundle adjustment Dense matching Data products

More information

ACCELERATED DUAL GRADIENT-BASED METHODS FOR TOTAL VARIATION IMAGE DENOISING/DEBLURRING PROBLEMS. Donghwan Kim and Jeffrey A.

ACCELERATED DUAL GRADIENT-BASED METHODS FOR TOTAL VARIATION IMAGE DENOISING/DEBLURRING PROBLEMS. Donghwan Kim and Jeffrey A. ACCELERATED DUAL GRADIENT-BASED METHODS FOR TOTAL VARIATION IMAGE DENOISING/DEBLURRING PROBLEMS Donghwan Kim and Jeffrey A. Fessler University of Michigan Dept. of Electrical Engineering and Computer Science

More information

David G. Luenberger Yinyu Ye. Linear and Nonlinear. Programming. Fourth Edition. ö Springer

David G. Luenberger Yinyu Ye. Linear and Nonlinear. Programming. Fourth Edition. ö Springer David G. Luenberger Yinyu Ye Linear and Nonlinear Programming Fourth Edition ö Springer Contents 1 Introduction 1 1.1 Optimization 1 1.2 Types of Problems 2 1.3 Size of Problems 5 1.4 Iterative Algorithms

More information

An Enhanced Primal Dual Method For Total Variation-Based Wavelet Domain Inpainting

An Enhanced Primal Dual Method For Total Variation-Based Wavelet Domain Inpainting An Enhanced Primal Dual Method For Total Variation-Based Wavelet Domain Inpainting P. Sravan Kumar 1, H.V. Ram Ravi Kishore 2, V. Ravi Kumar 3 1 MTech, Spatial Information Technology, from JNTU, research

More information

EE795: Computer Vision and Intelligent Systems

EE795: Computer Vision and Intelligent Systems EE795: Computer Vision and Intelligent Systems Spring 2012 TTh 17:30-18:45 FDH 204 Lecture 14 130307 http://www.ee.unlv.edu/~b1morris/ecg795/ 2 Outline Review Stereo Dense Motion Estimation Translational

More information

Optic Flow and Basics Towards Horn-Schunck 1

Optic Flow and Basics Towards Horn-Schunck 1 Optic Flow and Basics Towards Horn-Schunck 1 Lecture 7 See Section 4.1 and Beginning of 4.2 in Reinhard Klette: Concise Computer Vision Springer-Verlag, London, 2014 1 See last slide for copyright information.

More information

On-line and Off-line 3D Reconstruction for Crisis Management Applications

On-line and Off-line 3D Reconstruction for Crisis Management Applications On-line and Off-line 3D Reconstruction for Crisis Management Applications Geert De Cubber Royal Military Academy, Department of Mechanical Engineering (MSTA) Av. de la Renaissance 30, 1000 Brussels geert.de.cubber@rma.ac.be

More information

Action Recognition with HOG-OF Features

Action Recognition with HOG-OF Features Action Recognition with HOG-OF Features Florian Baumann Institut für Informationsverarbeitung, Leibniz Universität Hannover, {last name}@tnt.uni-hannover.de Abstract. In this paper a simple and efficient

More information

EFFICIENT AND GLOBALLY OPTIMAL MULTI VIEW DENSE MATCHING FOR AERIAL IMAGES

EFFICIENT AND GLOBALLY OPTIMAL MULTI VIEW DENSE MATCHING FOR AERIAL IMAGES EFFICIENT AND GLOBALLY OPTIMAL MULTI VIEW DENSE MATCHING FOR AERIAL IMAGES Arnold Irschara a, Markus Rumpler b, Philipp Meixner b, Thomas Pock b and Horst Bischof b a Microsoft Photogrammety, Anzengrubergasse

More information

Live Metric 3D Reconstruction on Mobile Phones ICCV 2013

Live Metric 3D Reconstruction on Mobile Phones ICCV 2013 Live Metric 3D Reconstruction on Mobile Phones ICCV 2013 Main Contents 1. Target & Related Work 2. Main Features of This System 3. System Overview & Workflow 4. Detail of This System 5. Experiments 6.

More information

Label Configuration Priors for Continuous Multi-Label Optimization. Mohamed Souiai, Evgeny Strekalovskiy, Claudia Nieuwenhuis, Daniel Cremers

Label Configuration Priors for Continuous Multi-Label Optimization. Mohamed Souiai, Evgeny Strekalovskiy, Claudia Nieuwenhuis, Daniel Cremers TUM TECHNISCHE UNIVERSITÄT MÜNCHEN INSTITUT FÜR INFORMATIK Label Configuration Priors for Continuous Multi-Label Optimization Mohamed Souiai, Evgeny Strekalovskiy, Claudia Nieuwenhuis, Daniel Cremers TUM-I1221

More information

Optimality Bounds for a Variational Relaxation of the Image Partitioning Problem

Optimality Bounds for a Variational Relaxation of the Image Partitioning Problem Optimality Bounds for a Variational Relaxation of the Image Partitioning Problem Jan Lellmann, Frank Lenzen, Christoph Schnörr Image and Pattern Analysis Group Universität Heidelberg EMMCVPR 2011 St. Petersburg,

More information

Optical Flow CS 637. Fuxin Li. With materials from Kristen Grauman, Richard Szeliski, S. Narasimhan, Deqing Sun

Optical Flow CS 637. Fuxin Li. With materials from Kristen Grauman, Richard Szeliski, S. Narasimhan, Deqing Sun Optical Flow CS 637 Fuxin Li With materials from Kristen Grauman, Richard Szeliski, S. Narasimhan, Deqing Sun Motion and perceptual organization Sometimes, motion is the only cue Motion and perceptual

More information

Comparison of stereo inspired optical flow estimation techniques

Comparison of stereo inspired optical flow estimation techniques Comparison of stereo inspired optical flow estimation techniques ABSTRACT The similarity of the correspondence problems in optical flow estimation and disparity estimation techniques enables methods to

More information

Motion Estimation (II) Ce Liu Microsoft Research New England

Motion Estimation (II) Ce Liu Microsoft Research New England Motion Estimation (II) Ce Liu celiu@microsoft.com Microsoft Research New England Last time Motion perception Motion representation Parametric motion: Lucas-Kanade T I x du dv = I x I T x I y I x T I y

More information

An Integrated System for Digital Restoration of Prehistoric Theran Wall Paintings

An Integrated System for Digital Restoration of Prehistoric Theran Wall Paintings An Integrated System for Digital Restoration of Prehistoric Theran Wall Paintings Nikolaos Karianakis 1 Petros Maragos 2 1 University of California, Los Angeles 2 National Technical University of Athens

More information

Nonlinear Programming

Nonlinear Programming Nonlinear Programming SECOND EDITION Dimitri P. Bertsekas Massachusetts Institute of Technology WWW site for book Information and Orders http://world.std.com/~athenasc/index.html Athena Scientific, Belmont,

More information

Multiplicative noise removal using primal dual and reweighted alternating minimization

Multiplicative noise removal using primal dual and reweighted alternating minimization Wang et al. SpringerPlus 06) 5:77 DOI 0.86/s40064-06-807-3 RESEARCH Open Access Multiplicative noise removal using primal dual and reweighted alternating minimization Xudong Wang *, Yingzhou Bi, Xiangchu

More information

Image denoising using TV-Stokes equation with an orientation-matching minimization

Image denoising using TV-Stokes equation with an orientation-matching minimization Image denoising using TV-Stokes equation with an orientation-matching minimization Xue-Cheng Tai 1,2, Sofia Borok 1, and Jooyoung Hahn 1 1 Division of Mathematical Sciences, School of Physical Mathematical

More information

Optical flow and depth from motion for omnidirectional images using a TV-L1 variational framework on graphs

Optical flow and depth from motion for omnidirectional images using a TV-L1 variational framework on graphs ICIP 2009 - Monday, November 9 Optical flow and depth from motion for omnidirectional images using a TV-L1 variational framework on graphs Luigi Bagnato Signal Processing Laboratory - EPFL Advisors: Prof.

More information

Convex Optimization for Scene Understanding

Convex Optimization for Scene Understanding Convex Optimization for Scene Understanding Mohamed Souiai 1, Claudia Nieuwenhuis 2, Evgeny Strekalovskiy 1 and Daniel Cremers 1 1 Technical University of Munich 2 UC Berkeley, ICSI, USA Abstract In this

More information

Optical flow. Cordelia Schmid

Optical flow. Cordelia Schmid Optical flow Cordelia Schmid Motion field The motion field is the projection of the 3D scene motion into the image Optical flow Definition: optical flow is the apparent motion of brightness patterns in

More information

The p-laplacian on Graphs with Applications in Image Processing and Classification

The p-laplacian on Graphs with Applications in Image Processing and Classification The p-laplacian on Graphs with Applications in Image Processing and Classification Abderrahim Elmoataz 1,2, Matthieu Toutain 1, Daniel Tenbrinck 3 1 UMR6072 GREYC, Université de Caen Normandie 2 Université

More information

Semantic 3D Reconstruction of Heads Supplementary Material

Semantic 3D Reconstruction of Heads Supplementary Material Semantic 3D Reconstruction of Heads Supplementary Material Fabio Maninchedda1, Christian Ha ne2,?, Bastien Jacquet3,?, Amae l Delaunoy?, Marc Pollefeys1,4 1 ETH Zurich 2 UC Berkeley 3 Kitware SAS 4 Microsoft

More information

TV-L 1 Optical Flow for Vector Valued Images

TV-L 1 Optical Flow for Vector Valued Images TV-L 1 Optical Flow for Vector Valued Images Lars Lau Rakêt 1, Lars Roholm 2, Mads Nielsen 1, and François Lauze 1 1 Department of Computer Science, University of Copenhagen, Denmark {larslau, madsn, francois}@diku.dk

More information

Static Scene Reconstruction

Static Scene Reconstruction GPU supported Real-Time Scene Reconstruction with a Single Camera Jan-Michael Frahm, 3D Computer Vision group, University of North Carolina at Chapel Hill Static Scene Reconstruction 1 Capture on campus

More information

Image Smoothing and Segmentation by Graph Regularization

Image Smoothing and Segmentation by Graph Regularization Image Smoothing and Segmentation by Graph Regularization Sébastien Bougleux 1 and Abderrahim Elmoataz 1 GREYC CNRS UMR 6072, Université de Caen Basse-Normandie ENSICAEN 6 BD du Maréchal Juin, 14050 Caen

More information

Stereo Vision II: Dense Stereo Matching

Stereo Vision II: Dense Stereo Matching Stereo Vision II: Dense Stereo Matching Nassir Navab Slides prepared by Christian Unger Outline. Hardware. Challenges. Taxonomy of Stereo Matching. Analysis of Different Problems. Practical Considerations.

More information

Multiple Frame Integration for OCR on Mobile Devices

Multiple Frame Integration for OCR on Mobile Devices Multiple Frame Integration for OCR on Mobile Devices Master s Thesis Georg Krispel Advisor: Horst Bischof December 12, 2016 Institute for Computer Graphics and Vision Anyline GmbH Scene Text Recognition

More information

Incremental Line-based 3D Reconstruction using Geometric Constraints

Incremental Line-based 3D Reconstruction using Geometric Constraints HOFER ET AL.: INCREMENTAL LINE-BASED 3D RECONSTRUCTION 1 Incremental Line-based 3D Reconstruction using Geometric Constraints Manuel Hofer hofer@icg.tugraz.at Andreas Wendel wendel@icg.tugraz.at Horst

More information

Semi-Dense Direct SLAM

Semi-Dense Direct SLAM Computer Vision Group Technical University of Munich Jakob Engel Jakob Engel, Daniel Cremers David Caruso, Thomas Schöps, Lukas von Stumberg, Vladyslav Usenko, Jörg Stückler, Jürgen Sturm Technical University

More information

An Introduction to Optimization Techniques in Computer Graphics

An Introduction to Optimization Techniques in Computer Graphics An Introduction to Optimization Techniques in Computer Graphics Ivo Ihrke, Xavier Granier, Gaël Guennebaud, Laurent Jacques, Bastian Goldluecke To cite this version: Ivo Ihrke, Xavier Granier, Gaël Guennebaud,

More information

Algorithms for Markov Random Fields in Computer Vision

Algorithms for Markov Random Fields in Computer Vision Algorithms for Markov Random Fields in Computer Vision Dan Huttenlocher November, 2003 (Joint work with Pedro Felzenszwalb) Random Field Broadly applicable stochastic model Collection of n sites S Hidden

More information

Sparse wavelet expansions for seismic tomography: Methods and algorithms

Sparse wavelet expansions for seismic tomography: Methods and algorithms Sparse wavelet expansions for seismic tomography: Methods and algorithms Ignace Loris Université Libre de Bruxelles International symposium on geophysical imaging with localized waves 24 28 July 2011 (Joint

More information

An Efficient Octree Design for Local Variational Range Image Fusion

An Efficient Octree Design for Local Variational Range Image Fusion An Efficient Octree Design for Local Variational Range Image Fusion Nico Marniok, Ole Johannsen, and Bastian Goldluecke University of Konstanz, Konstanz, Germany Abstract. We present a reconstruction pipeline

More information

Robust Optical Flow Computation Under Varying Illumination Using Rank Transform

Robust Optical Flow Computation Under Varying Illumination Using Rank Transform Sensors & Transducers Vol. 69 Issue 4 April 04 pp. 84-90 Sensors & Transducers 04 by IFSA Publishing S. L. http://www.sensorsportal.com Robust Optical Flow Computation Under Varying Illumination Using

More information