12.07.2015 Views

Metrics of curves in shape optimization and analysis - Andrea Carlo ...

Metrics of curves in shape optimization and analysis - Andrea Carlo ...

Metrics of curves in shape optimization and analysis - Andrea Carlo ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

6.2.4 Applications <strong>in</strong> computer visionCharpiat et al. [9] propose an approximation method to compute len H (ξ) bymeans <strong>of</strong> a family <strong>of</strong> energies def<strong>in</strong>ed us<strong>in</strong>g a smooth <strong>in</strong>tegr<strong>and</strong>; the approximationis ma<strong>in</strong>ly based on the property ‖f‖ L p → p ‖f‖ L ∞, for any measurable functionf def<strong>in</strong>ed on a bounded doma<strong>in</strong>; they successively devise a method to f<strong>in</strong>dapproximation <strong>of</strong> geodesics.6.3 A Hausdorff-like distance <strong>of</strong> compact setsIn Duci <strong>and</strong> Mennucci [15] a L p -like distance on the compact subsets <strong>of</strong> lR N wasproposed. (Here p ∈ [1, ∞).)To this end, we fix ϕ : [0, ∞) → (0, ∞) , decreas<strong>in</strong>g, C 1 , with ϕ(|x|) ∈ L p . Wethen def<strong>in</strong>e v A (x) := ϕ(u A (x)), where u A is the distance function. We eventuallydef<strong>in</strong>e the distanced(A, B) := ‖v A − v B ‖ L p . (6.2)Remark 6.15 This <strong>shape</strong> space is a perfect example <strong>of</strong> the representation/embedd<strong>in</strong>g/quotient scheme. Indeed, this <strong>shape</strong> space is represented as N c ={v A | A compact} <strong>and</strong> embedded <strong>in</strong> L p . Given v ∈ N c , we recover the <strong>shape</strong>A = {v = ϕ −1 (0)} (that is a level set <strong>of</strong> v).Example 6.16 A simple example (that works for all p) is given by ϕ(t) = e −t ,so that v A (x) = exp (−u A (x)); <strong>in</strong> this case, A = {v = 1}.The distance d <strong>of</strong> eqn. (6.2) enjoys the follow<strong>in</strong>g properties.Properties 6.17• It is Euclidean <strong>in</strong>variant;• it is locally compact but not path-metric;• the topology <strong>in</strong>duced is the same as that <strong>in</strong>duced by d H ;• m<strong>in</strong>imal geodesics do exist, s<strong>in</strong>ce D g := {B | d g (A, B) ≤ ρ} is compact.The pro<strong>of</strong>s are <strong>in</strong> [15]. We present a numerical computation <strong>of</strong> a m<strong>in</strong>imal geodesic(by A. Duci) <strong>in</strong> Figure 7 on the next page.6.3.1 Analogy with the Hausdorff metric, L p vs L ∞We recall that d H (A, B) = ‖u A (x) − u B (x)‖ L ∞ ; whereas <strong>in</strong>stead now we arepropos<strong>in</strong>g d(A, B) := ‖v A −v B ‖ L p . The idea be<strong>in</strong>g that this distance <strong>of</strong> compactsets is modeled on L p , whereas the Hausdorff distance is “modeled” on L ∞ .(Note that the Hausdorff distance is not really obta<strong>in</strong>ed by embedd<strong>in</strong>g, s<strong>in</strong>ceu A ∉ L ∞ ). L p is more regular than L ∞ , as shown by this remark.Remark 6.18 Given any f, g ∈ L p with p ∈ (1, ∞), the segment connect<strong>in</strong>gf to g is the unique m<strong>in</strong>imal geodesic connect<strong>in</strong>g them. Suppose now that thedimension <strong>of</strong> L ∞ (Ω, A, µ) is greater than 1. Given generically f, g ∈ L ∞ , thereis an uncountable number <strong>of</strong> m<strong>in</strong>imal geodesics connect<strong>in</strong>g them. 99 “Generically” is meant <strong>in</strong> the Baire sense: the set <strong>of</strong> exceptions is <strong>of</strong> first category.46

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!