1 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Invariant shape similarity © Alexander & Michael Bronstein, 2006-2009 © Michael Bronstein,

Slides:



Advertisements
Similar presentations
Differential geometry I
Advertisements

Spectral methods 1 © Alexander & Michael Bronstein,
Graph Laplacian Regularization for Large-Scale Semidefinite Programming Kilian Weinberger et al. NIPS 2006 presented by Aggeliki Tsoli.
Topology-Invariant Similarity and Diffusion Geometry
1 Numerical Geometry of Non-Rigid Shapes Diffusion Geometry Diffusion geometry © Alexander & Michael Bronstein, © Michael Bronstein, 2010 tosca.cs.technion.ac.il/book.
Discrete Geometry Tutorial 2 1
EARS1160 – Numerical Methods notes by G. Houseman
Xianfeng Gu, Yaling Wang, Tony Chan, Paul Thompson, Shing-Tung Yau
Isometry-Invariant Similarity
Discrete geometry Lecture 2 1 © Alexander & Michael Bronstein
1 Numerical geometry of non-rigid shapes Geometry Numerical geometry of non-rigid shapes Shortest path problems Alexander Bronstein, Michael Bronstein,
1 Numerical geometry of non-rigid shapes Consistent approximation of geodesics in graphs Consistent approximation of geodesics in graphs Tutorial 3 © Alexander.
Shape reconstruction and inverse problems
Fast marching methods The continuous way 1
Invariant correspondence
1 Processing & Analysis of Geometric Shapes Shortest path problems Shortest path problems The discrete way © Alexander & Michael Bronstein, ©
1 Michael Bronstein Computational metric geometry Computational metric geometry Michael Bronstein Department of Computer Science Technion – Israel Institute.
1 Numerical geometry of non-rigid shapes Partial similarity Partial similarity © Alexander & Michael Bronstein, © Michael Bronstein, 2010 tosca.cs.technion.ac.il/book.
1 Processing & Analysis of Geometric Shapes Introduction Processing and Analysis of Geometric Shapes Department of Electrical Engineering – Technion Spring.
Multidimensional scaling
Numerical Optimization
Isometry invariant similarity
Spectral embedding Lecture 6 1 © Alexander & Michael Bronstein
Fast marching methods Lecture 3 1 © Alexander & Michael Bronstein
Numerical geometry of non-rigid shapes
CPSC 689: Discrete Algorithms for Mobile and Wireless Systems Spring 2009 Prof. Jennifer Welch.
Numerical geometry of objects
Motion Analysis (contd.) Slides are from RPI Registration Class.
1 Bronstein 2 and Kimmel Extrinsic and intrinsic similarity of nonrigid shapes Michael M. Bronstein Department of Computer Science Technion – Israel Institute.
Lecture IV – Invariant Correspondence
1 Numerical geometry of non-rigid shapes Mathematical background Mathematical background Tutorial 1 © Maks Ovsjanikov, Alex & Michael Bronstein tosca.cs.technion.ac.il/book.
Correspondence & Symmetry
Constrained Optimization
1 Numerical geometry of non-rigid shapes Spectral Methods Tutorial. Spectral Methods Tutorial 6 © Maks Ovsjanikov tosca.cs.technion.ac.il/book Numerical.
1 Numerical geometry of non-rigid shapes Lecture II – Numerical Tools Numerical geometry of shapes Lecture II – Numerical Tools non-rigid Alex Bronstein.
Spectral Embedding Alexander Bronstein, Michael Bronstein
Numerical geometry of non-rigid shapes
1 Numerical geometry of non-rigid shapes In the Rigid Kingdom In the Rigid Kingdom Lecture 4 © Alexander & Michael Bronstein tosca.cs.technion.ac.il/book.
1 Numerical geometry of non-rigid shapes A journey to non-rigid world objects Invariant correspondence and shape synthesis non-rigid Alexander Bronstein.
1 Bronstein, Bronstein, and Kimmel Joint extrinsic and intrinsic similarity of non-rigid shapes Rock, paper, and scissors Joint extrinsic and intrinsic.
Invariant Correspondence
1 Regularized partial similarity of shapes NORDIA – CVPR 2008 Not only size matters: Regularized partial similarity of shapes Alexander Bronstein, Michael.
1 Numerical geometry of non-rigid shapes A journey to non-rigid world objects Numerical methods non-rigid Alexander Bronstein Michael Bronstein Numerical.
Non-Euclidean Embedding
1 Numerical geometry of non-rigid shapes Numerical Geometry Numerical geometry of non-rigid shapes Numerical geometry Alexander Bronstein, Michael Bronstein,
Numerical geometry of non-rigid shapes
Paretian similarity for partial comparison of non-rigid objects
1 Bronstein 2 & Kimmel An isometric model for facial animation and beyond AMDO, Puerto de Andratx, 2006 An isometric model for facial animation and beyond.
1 Numerical geometry of non-rigid shapes Non-Euclidean Embedding Non-Euclidean Embedding Lecture 6 © Alexander & Michael Bronstein tosca.cs.technion.ac.il/book.
1 Michael M. Bronstein Partial similarity of objects 17 December 2006 Partial similarity of objects, or how to compare a centaur to a horse Michael M.
1 Bronstein 2 & Kimmel Matching 2D articulated shapes using GMDS AMDO, Puerto de Andratx, 2006 Matching 2D articulated shapes using Generalized Multidimensional.
1 Numerical geometry of non-rigid shapes Shortest Path Problems Shortest Path Problems Alexander Bronstein, Michael Bronstein © 2008 All rights reserved.
CSci 6971: Image Registration Lecture 5: Feature-Base Regisration January 27, 2004 Prof. Chuck Stewart, RPI Dr. Luis Ibanez, Kitware Prof. Chuck Stewart,
1 M. Bronstein Multigrid multidimensional scaling Multigrid Multidimensional Scaling Michael M. Bronstein Department of Computer Science Technion – Israel.
1 Numerical geometry of non-rigid shapes Non-rigid correspondence Numerical geometry of non-rigid shapes Non-rigid correspondence Alexander Bronstein,
1 Numerical geometry of non-rigid shapes Nonrigid Correspondence & Calculus of Shapes Non-Rigid Correspondence and Calculus of Shapes Of bodies changed.
Intrinsic Parameterization for Surface Meshes Mathieu Desbrun, Mark Meyer, Pierre Alliez CS598MJG Presented by Wei-Wen Feng 2004/10/5.
CSE554Laplacian DeformationSlide 1 CSE 554 Lecture 8: Laplacian Deformation Fall 2012.
Surface Simplification Using Quadric Error Metrics Michael Garland Paul S. Heckbert.
1 Numerical geometry of non-rigid shapes Shortest path problems Shortest path problems Lecture 2 © Alexander & Michael Bronstein tosca.cs.technion.ac.il/book.
Coarse-to-Fine Combinatorial Matching for Dense Isometric Shape Correspondence Yusuf Sahillioğlu and Yücel Yemez Computer Eng. Dept., Koç University, Istanbul,
Expression-invariant Face Recognition using Geodesic Distance Isometries Kerry Widder A Review of ‘Robust expression-invariant face recognition from partially.
Course 13 Curves and Surfaces. Course 13 Curves and Surface Surface Representation Representation Interpolation Approximation Surface Segmentation.
Computer Graphics Some slides courtesy of Pierre Alliez and Craig Gotsman Texture mapping and parameterization.
Introduction to Optimization
1 Numerical geometry of non-rigid shapes Projects Quasi-isometries Project 1 © Alexander & Michael Bronstein tosca.cs.technion.ac.il/book Numerical geometry.
Approximation Algorithms based on linear programming.
Morphing and Shape Processing
Spectral Methods Tutorial 6 1 © Maks Ovsjanikov
Craig Schroeder October 26, 2004
Presentation transcript:

1 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Invariant shape similarity © Alexander & Michael Bronstein, © Michael Bronstein, 2010 tosca.cs.technion.ac.il/book Advanced topics in vision Processing and Analysis of Geometric Shapes EE Technion, Spring 2010

2 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Invariant similarity SIMILARITY TRANSFORMATION

3 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Equivalence Equal CongruentIsometric

4 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Equivalence Equivalence is a binary relation on the space of shapes which for all satisfies Reflexivity: Symmetry: Transitivity: Can be expressed as a binary function if and only if Quotient space is the space of equivalence classes

5 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Equivalence

6 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Equivalence All deformations of the human shape are “the same”

7 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Similarity Shapes are rarely truly equivalent (e.g., due to acquisition noise or since most shapes are rigid) We want to account for “almost equivalence” or similarity -similar = -isometric (w.r.t. some metric) Define a distance on the shape space quantifying the degree of dissimilarity of shapes

8 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Similarity A monkey shape is more similar to a deformation of a monkey shape… …than to a human shape

9 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Isometry-invariant distance Non-negative function satisfying for all Corollary: is a metric on the quotient space Similarity: and are -isometric; and are -isometric (In particular, if and only if ) Symmetry: Triangle inequality: Given discretized shapes and sampled with radius Consistency to sampling:

10 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Canonical forms distance Compute Hausdorff distance over all isometries in Minimum-distortion embedding Minimum-distortion embedding No fixed embedding space will give distortion-less canonical forms

11 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Gromov-Hausdorff distance Isometric embedding Isometric embedding Mikhail Gromov Gromov-Hausdorff distance: include into minimization

12 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Properties of Gromov-Hausdorff distance Metric on the quotient space of isometries of shapes Similarity: and are -isometric; and are -isometric Consistent to sampling: given discretized shapes and sampled with radius Generalization of Hausdorff distance: Hausdorff distance between subsets of a metric space Gromov-Hausdorff distance between metric spaces Gromov, 1981

13 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Alternative definition I (metric coupling) is the disjoint union of and the (semi-) metric satisfies and where Mémoli, 2008

14 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Optimization over translates into finding the values of Mémoli, 2008 A lot of constraints! Alternative definition I (metric coupling)

15 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Correspondence A subset is called a correspondence between and if for every there exists at least one such that and similarly for every there exists such that Particular case: given and

16 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Correspondence distortion The distortion of correspondence is defined as

17 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Alternative definition II (correspondence distortion) 1. Show that for any there exists with Proof sketch Since, by definition of, and are subspaces of some such that Let By triangle inequality, for

18 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Alternative definition II (correspondence distortion) 2. Show that for any Let It is sufficient to show that there is a (semi-)metric on the disjoint union such that,, and Construct the metric as follows (in particular, for ).

19 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Alternative definition II (correspondence distortion) First, For each Since for, Second, we need to show that is a (semi-)metric on On and, it is straightforward We only need to show metric properties hold on

20 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Alternative definition III measures how much is distorted by when embedded into

21 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity measures how much is distorted by when embedded into Alternative definition III

22 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity measures how far is from being the inverse of Alternative definition III

23 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Generalized MDS A. Bronstein, M. Bronstein & R. Kimmel, 2006

24 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Difficulties How to represent points on ? Global parametrization is not always available. Some local representation is required in general case. No more closed-form expression for. Metric needs to be approximated. Minimization algorithm.

25 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Local representation is sampled at and represented as a triangular mesh. Any point falls into one of the triangles. Within the triangle, it can be represented as convex combination of triangle vertices, Barycentric coordinates. We will need to handle discrete indices in minimization algorithm.

26 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Geodesic distances Distance terms can be precomputed, since are fixed. How to compute distance terms ? No more closed-form expression. Cannot be precomputed, since are minimization variables. can fall anywhere on the mesh. Precompute for all. Approximate for any.

27 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Geodesic distance approximation Approximation from. First order accurate: Consistent with data: Symmetric: Smoothness: is and a closed-form expression for its derivatives is available to minimization algorithm. Might be only at some points or along some lines. Efficiently computed: constant complexity independent of.

28 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Geodesic distance approximation Compute for. falls into triangle and is represented as Particular case: Hence, we can precompute distances How to compute from ?

29 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Geodesic distance approximation We have already encountered this problem in fast marching. Wavefront arrives at triangle vertex at time. When does it arrive to ? Adopt planar wavefront model. Distance map is linear in the triangle (hence, linear in ) Solve for coefficients and obtain a linear interpolant

30 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Geodesic distance approximation General case: falls into triangle and is represented as Apply previous steps in triangle to obtain Apply once again in triangle to obtain

31 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity A four-step dance

32 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Minimization algorithm How to minimize the generalized stress? Particular case: L 2 stress Fix all and all except for some. Stress as a function of only becomes quadratic

33 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Quadratic stress is positive semi-definite. is convex in (but not necessarily in together).

34 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Quadratic stress Closed-form solution for minimizer of Problem: solution might be outside the triangle. Solution: find constrained minimizer Closed-form solution still exists.

35 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Minimization algorithm Initialize For each Fix and compute gradient Select corresponding to maximum. Compute minimizer If constraints are active translate to adjacent triangle. Iterate until convergence…

36 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity How to move to adjacent triangles? Three cases All : inside triangle. : on edge opposite to. : on vertex. inside on edgeon vertex

37 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Point on edge on edge opposite to. If edge is not shared by any other triangle we are on the boundary – no translation. Otherwise, express the point as in triangle. contains same values as. May be permuted due to different vertex ordering in. Complication: is not on the edge. Evaluate gradient in. If points inside triangle, update to.

38 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Point on vertex on vertex. For each triangle sharing vertex Express point as in. Evaluate gradient in. Reject triangles with pointing outside. Select triangle with maximum. Update to.

39 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity MDS vs GMDS Stress Generalized MDSMDS Generalized stress Analytic expression for Nonconvex problem Variables: Euclidean coordinates of the points must be interpolated Nonconvex problem Variables: points on in barycentric coordinates

40 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Multiresolution Stress is non convex – many small local minima. Straightforward minimization gives poor results. How to initialize GMDS? Multiresolution: Create a hierarchy of grids in, Each grid comprises Sampling: Geodesic distance matrix:

41 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Multiresolution Initialize at the coarsest resolution in. For Starting at initialization, solve the GMDS problem Interpolate solution to next resolution level Return.

42 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity GMDS Interpolation GMDS

43 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Multiresolution encore So far, we created a hierarchy of embedded spaces. One step further: create a hierarchy of embedding spaces.

44 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Labeling problem Torresani, Kolmogorov, Rother 2008 Wang, B 2010 Build a graph with vertices and edges Label each vertex Minimum distortion correspondence = graph labeling problem Efficient solvers with good global convergence properties Complexity: Hierarchical solution complexity can be lowered to

45 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity MATLAB ® intermezzo GMDS

46 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Discrete Gromov-Hausdorff distance Two coupled GMDS problems Can be cast as a constrained problem Bronstein, Bronstein & Kimmel, 2006

47 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Numerical example Canonical forms (MDS based on 500 points) Gromov-Hausdorff distance (GMDS based on 50 points) Bronstein, Bronstein & Kimmel, 2006

48 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Extrinsic similarity using Gromov-Hausdorff distance Mémoli (2008) Connection between Euclidean GH and ICP distances: CongruenceEuclidean isometry EXTRINSIC SIMILARITY ICP distance: GH distance with Euclidean metric: Mémoli, 2008

49 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Connection to canonical form distance

50 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Correspondence again A different representation for correspondence using indicator functions defines a valid correspondence if

51 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity L p Gromov-Hausdorff distance We can give an alternative formulation of the Gromov-Hausdorff distance Can we define an L p version of the Gromov-Hausdorff distance by relaxing the above definition?

52 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Measure coupling Let be probability measures defined on and The measure can be considered as a relaxed version of the indicator function or as fuzzy correspondence A measure on is a coupling of and if for all measurable sets Mémoli, 2007 (a metric space with measure is called a metric measure or mm space)

53 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Gromov-Wasserstein distance The relaxed version of the Gromov-Hausdorff distance is given by and is referred to as Gromov-Wasserstein distance [Memoli 2007] Mémoli, 2007

54 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity Earth Mover’s distance Let be a metric space, and measures supported on EMD as optimal mass transport: mass transported from to distance traveled Mémoli, 2007 The Wasserstein or Earth Mover’s distance (EMD) is given by Define the coupling of

55 Numerical Geometry of Non-Rigid Shapes Invariant shape similarity The analogy Hausdorff Mémoli, 2007 Distance between subsets of a metric space. Gromov-Hausdorff Distance between metric spaces Wasserstein Distance between subsets of a metric measure space. Gromov-Wasserstein Distance between metric measure spaces