Smooth Shape-Aware Functions with Controlled Extrema Alec Jacobson 1 1 ETH Zurich Tino Weinkauf 2 2 MPI Saarbrücken Olga Sorkine 1 August 9, 2012
Real-time deformation relies on smooth, shape-aware functions input shape + handles August 9, 2012 Alec Jacobson # 2
Real-time deformation relies on smooth, shape-aware functions precompute weight functions August 9, 2012 Alec Jacobson # 3
Real-time deformation relies on smooth, shape-aware functions deform handles à deform shape August 9, 2012 Alec Jacobson # 4
Real-time deformation relies on smooth, shape-aware functions August 9, 2012 Alec Jacobson # 5
Real-time deformation relies on smooth, shape-aware functions August 9, 2012 Alec Jacobson # 6
Spurious extrema cause distracting artifacts unconstrained [Botsch & Kobbelt 2004] local max local min August 9, 2012 Alec Jacobson # 7
Spurious extrema cause distracting artifacts unconstrained [Botsch & Kobbelt 2004] local max local min August 9, 2012 Alec Jacobson # 8
Bounds help, but don’t solve problem bounded [Jacobson et al. 2011] local max local min August 9, 2012 Alec Jacobson # 9
Bounds help, but don’t solve problem bounded [Jacobson et al. 2011] local max local min August 9, 2012 Alec Jacobson # 10
Gets worse with higher-order smoothness bounded [Jacobson et al. 2011] local max local min oscillate too much August 9, 2012 Alec Jacobson # 11
Gets worse with higher-order smoothness bounded [Jacobson et al. 2011] local max local min oscillate too much August 9, 2012 Alec Jacobson # 12
We explicitly prohibit spurious extrema our local max local min August 9, 2012 Alec Jacobson # 13
We explicitly prohibit spurious extrema our local max local min August 9, 2012 Alec Jacobson # 14
Same functions used for color interpolation August 9, 2012 Alec Jacobson # 15
Same functions used for color interpolation August 9, 2012 Alec Jacobson # 16
Same functions used for color interpolation unconstrained [Finch et al. 2011] Image courtesy Mark Finch August 9, 2012 Alec Jacobson # 17
Same functions used for color interpolation unconstrained [Finch et al. 2011] August 9, 2012 Alec Jacobson # 18
Same functions used for color interpolation unconstrained [Finch et al. 2011] Our August 9, 2012 Alec Jacobson # 19
Want same control when smoothing data August 9, 2012 Alec Jacobson # 20
Want same control when smoothing data Exact, but sharp geodesic August 9, 2012 Alec Jacobson # 21
Want same control when smoothing data Exact, but sharp geodesic August 9, 2012 Alec Jacobson # 22
Want same control when smoothing data Exact, but sharp geodesic Smooth, but extrema are lost August 9, 2012 Alec Jacobson # 23
Want same control when smoothing data Exact, but sharp geodesic Smooth and maintain extrema August 9, 2012 Alec Jacobson # 24
Ideal discrete problem is intractable arg min E ( f ) f Interpolation functions: August 9, 2012 Alec Jacobson # 25
Ideal discrete problem is intractable arg min E ( f ) f Data smoothing: August 9, 2012 Alec Jacobson # 26
Ideal discrete problem is intractable arg min E ( f ) f August 9, 2012 Alec Jacobson # 27
Ideal discrete problem is intractable arg min E ( f ) f s.t. f max = known f min = known August 9, 2012 Alec Jacobson # 28
Ideal discrete problem is intractable arg min E ( f ) f s.t. f max = known f min = known f j < f max linear f max f j > f min f j August 9, 2012 Alec Jacobson # 29
Ideal discrete problem is intractable arg min E ( f ) f s.t. f max = known f min = known f j f j < f max linear f j > f min f i f i > min j ∈ N ( i ) f j nonlinear f i < max j ∈ N ( i ) f j August 9, 2012 Alec Jacobson # 30
Assume we have a feasible solution “Representative function” u u j < u max linear handles u j > u min u i > min j ∈ N ( i ) u j nonlinear interior u i < max j ∈ N ( i ) u j August 9, 2012 Alec Jacobson # 31
Assume we have a feasible solution “Representative function” u u j < u max handles u j > u min u i > min j ∈ N ( i ) u j interior u i < max j ∈ N ( i ) u j August 9, 2012 Alec Jacobson # 32
Copy “monotonicity” of representative arg min E ( f ) f s.t. f max = known f min = known ( f i − f j )( u i − u j ) > 0 ∀ ( i, j ) ∈ E linear At least one edge in either direction per vertex August 9, 2012 Alec Jacobson # 33
Rewrite as conic optimization Conic QP Optimize with MOSEK August 9, 2012 Alec Jacobson # 34
We always have harmonic representative 1 Z kr u k 2 dV arg min 2 Ω u August 9, 2012 Alec Jacobson # 35
We always have harmonic representative 1 Z kr u k 2 dV arg min 2 Ω u u max = 1 s.t. August 9, 2012 Alec Jacobson # 36
We always have harmonic representative 1 Z kr u k 2 dV arg min 2 Ω u u max = 1 s.t. u min = 0 s.t. August 9, 2012 Alec Jacobson # 37
We always have harmonic representative 1 Z kr u k 2 dV arg min 2 Ω u u max = 1 s.t. u min = 0 s.t. Works well when no input function exists August 9, 2012 Alec Jacobson # 38
Data energy may fight harmonic representative Anisotropic input data August 9, 2012 Alec Jacobson # 39
Data energy may fight harmonic representative Anisotropic input data Harmonic representative August 9, 2012 Alec Jacobson # 40
Data energy may fight harmonic representative Anisotropic input data Harmonic representative August 9, 2012 Alec Jacobson # 41
Data energy may fight harmonic representative Anisotropic input data Harmonic representative August 9, 2012 Alec Jacobson # 42
Data energy may fight harmonic representative Anisotropic input data Resulting solution with large August 9, 2012 Alec Jacobson # 43
If data exists, copy topology, too Anisotropic input data [Weinkauf et al. 2010] representative August 9, 2012 Alec Jacobson # 44
If data exists, copy topology, too Anisotropic input data Resulting solution with large August 9, 2012 Alec Jacobson # 45
Final algorithm is simple and efficient ● Data smoothing : topology-aware representative § Morse-smale + linear solve ~milliseconds August 9, 2012 Alec Jacobson # 46
Final algorithm is simple and efficient ● Data smoothing : topology-aware representative § Morse-smale + linear solve ~milliseconds ● Interpolation : harmonic representative § Linear solve ~milliseconds August 9, 2012 Alec Jacobson # 47
Final algorithm is simple and efficient ● Data smoothing : topology-aware representative § Morse-smale + linear solve ~milliseconds ● Interpolation : harmonic representative § Linear solve ~milliseconds ● Conic optimization § 2D ~milliseconds, 3D ~seconds August 9, 2012 Alec Jacobson # 48
Final algorithm is simple and efficient ● Data smoothing : topology-aware representative § Morse-smale + linear solve ~milliseconds ● Interpolation : harmonic representative § Linear solve ~milliseconds ● Conic optimization § 2D ~milliseconds, 3D ~seconds Interpolation: functions are precomputed August 9, 2012 Alec Jacobson # 49
We preserve troublesome appendages Bounded Our August 9, 2012 Alec Jacobson # 50
We preserve troublesome appendages Bounded Our August 9, 2012 Alec Jacobson # 51
We preserve troublesome appendages Bounded Our #
Our weights attach appendages to body [Botsch & Kobbelt 2004, Our method Jacobson et al. 2011] August 9, 2012 Alec Jacobson # 53
Extrema glue appendages to far-away handles [Botsch & Kobbelt 2004, Jacobson et al. 2011] August 9, 2012 Alec Jacobson # 54
Extrema glue appendages to far-away handles [Botsch & Kobbelt 2004, Jacobson et al. 2011] August 9, 2012 Alec Jacobson # 55
Our weights attach appendages to body Our method August 9, 2012 Alec Jacobson # 56
Our weights attach appendages to body Our method August 9, 2012 Alec Jacobson # 57
Extrema distort small features Unconstrained [Botsch & Kobbelt 2004] weight of middle point August 9, 2012 Alec Jacobson # 58
Extrema distort small features Unconstrained [Botsch & Kobbelt 2004] weight of middle point August 9, 2012 Alec Jacobson # 59
Extrema distort small features Bounded [Jacobson et al. 2011] weight of middle point August 9, 2012 Alec Jacobson # 60
“Monotonicity” helps preserve small features Bounded [Jacobson et al. 2011] Our August 9, 2012 Alec Jacobson # 61
Spurious extrema are unstable, may “flip” slightly larger region August 9, 2012 Alec Jacobson # 62
Spurious extrema are unstable, may “flip” slightly larger region August 9, 2012 Alec Jacobson # 63
Spurious extrema are unstable, may “flip” Unconstrained [Botsch & Kobbelt, 2004] #
Spurious extrema are unstable, may “flip” Unconstrained [Botsch & Kobbelt, 2004] #
Recommend
More recommend