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 ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
10.1.1 Related worksA family <strong>of</strong> metrics similar to H j above (but for the length dependent scalefactors 15 ) was studied (<strong>in</strong>dependently) <strong>in</strong> Michor <strong>and</strong> Mumford [36]: the Sobolevtypeweak Riemannian metric on Imm(S 1 , R 2 )∫〈h, k〉 Gjc=j∑〈Dsh, i Dsk〉 i ds ;c i=0<strong>in</strong> that paper the geodesic equation, horizontality, conserved momenta, lower <strong>and</strong>upper bounds on the <strong>in</strong>duced distance, <strong>and</strong> scalar curvatures are computed. Notethat this metric is locally equivalent to the above metrics def<strong>in</strong>ed <strong>in</strong> equation(10.2), (10.3).Charpiat et al <strong>in</strong> [10, 11] studied (aga<strong>in</strong> <strong>in</strong>dependently) some generalizedmetrics <strong>and</strong> relative gradient flows; <strong>in</strong> particular they def<strong>in</strong>ed the Sobolev-typemetric∫〈h 1 , h 2 〉 + 〈D s h 1 , D s h 2 〉 ds .10.1.2 Properties <strong>of</strong> H j metricscThis is a list <strong>of</strong> important properties that will be discussed <strong>in</strong> the follow<strong>in</strong>gsections.• Flow regularization: Sobolev gradient flows are smoother than H 0 flows.• PDE order reduc<strong>in</strong>g property: Sobolev gradient flows are lower order thanH 0 flows.• SAC does not require derivatives <strong>of</strong> the curve to be def<strong>in</strong>ed for manycommonly used region-based <strong>and</strong> edge-based energies.• Coarse-to-f<strong>in</strong>e deformation property: a SAC automatically favors coarsescaledeformations before mov<strong>in</strong>g to f<strong>in</strong>e-scale deformations; this is idealfor visual track<strong>in</strong>g.• Sobolev-type norms <strong>in</strong>duce a well-def<strong>in</strong>ed distance on space <strong>of</strong> <strong>curves</strong>;• moreover the structure <strong>of</strong> the completion <strong>of</strong> the space <strong>of</strong> immersed <strong>curves</strong>w.r.to H 1 <strong>and</strong> H 2 norm is fairly well understood. So they <strong>of</strong>fer a consistenttheory <strong>of</strong> <strong>shape</strong> <strong>optimization</strong> <strong>and</strong> <strong>shape</strong> <strong>analysis</strong>.10.2 Mathematical propertiesWe start summariz<strong>in</strong>g the ma<strong>in</strong> mathematical properties, that were presented <strong>in</strong>[35] mostly. We first <strong>of</strong> all cite this lemma.15 Though, a scale-<strong>in</strong>variant Sobolev metric is proposed <strong>in</strong> [36] <strong>in</strong> §4.8 as a sensible generalization.62