作者: Mathias Gallardo , Daniel Pizarro , Toby Collins , Adrien Bartoli
DOI: 10.1007/S11263-019-01214-Z
关键词: Image (mathematics) 、 Partial differential equation 、 Reconstruction method 、 Critical point (mathematics) 、 Surface (mathematics) 、 Hidden Markov model 、 Mathematical analysis 、 Deformation (mechanics) 、 Mathematics 、 Pattern recognition (psychology)
摘要: Shape-from-Template (SfT) is the problem of using a shape template to infer deformable object observed in an image. The usual case SfT ‘Surface’ SfT, where 2D surface embedded 3D, and image perspective projection. We introduce ‘Curve’ comprising two new cases 1D curve. first when curve second 3D present thorough theoretical study these for isometric deformations, which are good approximation ropes, cables wires. Unlike Surface we show that Curve only ever solvable up discrete ambiguities. necessary sufficient conditions solvability with critical point analysis. further unlike cannot be solved locally exact non-holonomic Partial Differential Equations. Our main technical contributions two-fold. First, give stable, global reconstruction method models as Hidden Markov Model. This can generate all candidate solutions. Second, non-convex refinement novel angle-based deformation parameterization. quantitative qualitative results showing real shaped objects such necklace successfully reconstructed SfT.