1 / 40

SIGGRAPH 2010

SIGGRAPH 2010. “ Spectral Mesh Processing” Bruno Lévy and Richard Hao Zhang. Spectral Mesh Processing Applications 2/2. Bruno Lévy INRIA - ALICE. Overview. 1D parameterization Surface quadrangulation Surface parameterization Surface characterization Green function Heat kernel.

early
Download Presentation

SIGGRAPH 2010

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. SIGGRAPH 2010 “Spectral Mesh Processing” Bruno Lévy and Richard Hao Zhang

  2. Spectral Mesh Processing Applications 2/2 Bruno Lévy INRIA - ALICE

  3. Overview • 1D parameterization • Surface quadrangulation • Surface parameterization • Surface characterization Green function Heat kernel 

  4. 1D surface parameterizationGraph Laplacian ai,j = wi,j > 0 if (i,j) is an edge ai,i = -S ai,j (1,1 … 1) is an eigenvector assoc. with 0 The second eigenvector is interresting [Fiedler 73, 75]

  5. 1D surface parameterizationFiedler vector Reorder with Fiedler vector FEM matrix, Non-zero entries

  6. 1D surface parameterizationFiedler vector Streaming meshes [Isenburg & Lindstrom]

  7. 1D surface parameterizationFiedler vector Streaming meshes [Isenburg & Lindstrom]

  8. F(u) = ½ ut A u Minimize 1D surface parameterizationFiedler vector F(u) = S wij (ui - uj)2

  9. F(u) = ½ ut A u Minimize 1D surface parameterizationFiedler vector F(u) = S wij (ui - uj)2 How to avoid trivial solution ? Constrained vertices ?

  10. F(u) = ½ ut A u Minimize 1D surface parameterizationFiedler vector F(u) = S wij (ui - uj)2 S ui = 0 subject to Global constraints are more elegant !

  11. F(u) = ½ ut A u Minimize 1D surface parameterizationFiedler vector F(u) = S wij (ui - uj)2 S ui = 0 subject to ½ S ui2 = 1 Global constraints are more elegant ! We need also to constrain the second mementum

  12. F(u) = ½ ut A u Minimize u L = A u - l11 - l2 u l1L = ut1 l2L = ½(ut u – 1) 1D surface parameterizationFiedler vector F(u) = S wij (ui - uj)2 S ui = 0 subject to ½ S ui2 = 1 L(u) = ½ ut A u - l1 ut1 - l2½ (utu - 1) u = eigenvector of A l1 = 0 l2 = eigenvalue

  13. 1D surface parameterizationFiedler vector Rem: Fiedler vector is also a minimizer of the Rayleigh quotient R(A,x) = xt A x xt x The other eigenvectors xi are the solutions of : minimize R(A,xi) subject to xit xj = 0 for j < i

  14. Overview • 1D parameterization • Surface quadrangulation • Surface parameterization • Surface characterization Green function Heat kernel 

  15. Surface quadrangulation Nodal sets are sets of curves intersecting at constant angles The N-th eigenfunction has at most N eigendomains

  16. Surface quadrangulation

  17. Surface quadrangulation [L 2006], [Vallet & L 2006]

  18. Surface quadrangulation Filtered morse complex One eigenfunction Morse complex [Dong and Garland 2006]

  19. Surface quadrangulation Reparameterization of the quads

  20. Surface quadrangulation Improvement in [Huang, Zhang, Ma, Liu, Kobbelt and Bao 2008], takes a guidance vector field into account.

  21. Overview • 1D parameterization • Surface quadrangulation • Surface parameterization • Surface characterization Green function Heat kernel

  22. Surface parameterization S  v 2  u Discrete conformal mapping: [L, Petitjean, Ray, Maillot 2002] [Desbrun, Alliez 2002] - x y Minimize -  u  v T x y

  23. Surface parameterization S  v 2  u Discrete conformal mapping: [L, Petitjean, Ray, Maillot 2002] [Desbrun, Alliez 2002] - x y Minimize -  u  v T x y Uses pinned points.

  24. Surface parameterization [Muellen, Tong, Alliez, Desbrun 2008] Use Fiedler vector, i.e. the minimizer of R(A,x) = xt A x / xt x that is orthogonal to the trivial constant solution Implementation: (1) assemble the matrix of the discrete conformal parameterization (2) compute its eigenvector associated with the first non-zero eigenvalue See http://alice.loria.fr/WIKI/ Graphite tutorials – Manifold Harmonics

  25. Overview • 1D parameterization • Surface quadrangulation • Surface parameterization • Surface characterization Green function Heat kernel 

  26. Surface characterization Green Function Solving Poisson equation: f = g f = G(x,y)f(y) dy Where G: Green function is defined by: G(x,y) = d(x-y) d : dirac

  27. Surface characterization Green Function Solving Poisson equation: f = g f = G(x,y)f(y) dy Where G: Green function is defined by: G(x,y) = d(x-y) d : dirac Proof:   G(x,y)g(y) dy = d(x-y)g(y)dy = g(x) =  f(x)

  28. Surface characterization Green Function Solving Poisson equation: f = g f = G(x,y)f(y) dy Where G: Green function is defined by: G(x,y) = d(x-y) d : dirac Proof:   G(x,y)g(y) dy = d(x-y)g(y)dy = g(x) =  f(x)  f(x) = g(x) =   G(x,y)g(y)dy=  (G(x,y)g(y)dy)

  29. Surface characterization Green Function Solving Poisson equation: f = g f = G(x,y)f(y) dy Where G: Green function is defined by: G(x,y) = d(x-y) d : dirac Proof:   G(x,y)g(y) dy = d(x-y)g(y)dy = g(x) =  f(x)  f(x) = g(x) =   G(x,y)g(y)dy=  (G(x,y)g(y)dy) f(x) = G(x,y)g(y)dy

  30. Surface characterization Green Function How to compute G ? G is defined by: G(x,y) = d(x-y) d : dirac d(x-y) =  i(x) i(y) (completeness of the eigen decomposition)

  31. Surface characterization Green Function How to compute G ? G is defined by: G(x,y) = d(x-y) d : dirac d(x-y) =  i(x) i(y) (completeness of the eigen decomposition)  i(x) i(y) Works ! Using G(x,y) = li (G(x,y) =  i(x) i(y) = d(x-y) ) Note: Convergence of G series needs to be proved (complicated)

  32. Surface characterization Green Function  i(x) i(y) G(x,y) = li Summary: Solution Poisson equation: f = g f = G(x,y)f(y) dy Pose-invariant embedding Connection with GPS embedding [Rustamov 2007] GPS(x) = [1(x)/l1 , 2(x)/l2 , … , i(x)/li , … ] G(x,y) = GPS(x) * GPS(y)

  33. Overview • 1D parameterization • Surface quadrangulation • Surface parameterization • Surface characterization Green function Heat kernel 

  34. Surface characterization Heat equation f(x) (heat) The heat equation:  f = - f  t x t = 0

  35. Surface characterization Heat equation f(x) (heat) The heat equation:  f = - f  t x t = 100

  36. Surface characterization Heat equation f(x) (heat) The heat equation:  f = - f  t x t = 1000

  37. Surface characterization Heat equation Heat kernel: K(t,x,y) =  e-lit i(x) i(y) The heat equation:  f Solution of the heat equation: f(t,x) =  K(t,x,y) f(0,y) dy = - f  t

  38. Surface characterization Heat equation - What is the meaning of K(t,x,y) ? How much heat do we get at y after t seconds ? K(t,x,y) Initial time: we inject a Dirac of heat at x y y x x

  39. Surface characterization Heat equation Heat Kernel Signature [Sun, Ovsjanikov and Gubas 09] Auto-diffusion [Gebal, Baerentzen, Anaes and Larsen 09] How much heat remains at x after t seconds ? ADF(t,x) = HKS(t,x) = K(t,x,x) =  e-lit i2(x) Applications: shape signature, segmentation using Morse decomposition, …

  40. Summary • Minimizing Rayleigh quotient instead of using « pinning » • enforces global constraints (moments) that avoid the trivial solution • fieldler vector for streaming meshes [Isenburg et.al] • Spectral conformal parameterization [Muellen et.al] • The notion of fundamental solution plays a … fundamental role. • Strong connections with spectral analysis (and this is what Fourier • invented Fourier analysis for !) • Green function / Poisson equation - GPS coordinates [Rustamov] • Heat kernel signature ,[Sun et.al] / auto-diffusion function [Gebal et.al] • More to explore: the Variational Principle (see Wikipedia)

More Related