Vector field reconstruction from sparse samples with applications Marcos Lage, Fabiano Petronetto, Afonso Paiva, Hélio Lopes, Thomas Lewiner and Geovan Tavares Laboratório Matmidia - Departamento de Matemática - PUC-Rio - Rio de Janeiro - Brasil 1/23 Vector fields 2/23 Problem description 3/23 Problem description ? on the whole region 4/23 Motivation Particle Image Velocimetry (Luiz Fernando A. Azevedo, PUC-Rio) 5/23 Motivation Smoothed Particle Hydrodynamics (SPH) (F. Petronetto, A. Paiva, T. Lewiner, G. Tavares – PUC-Rio – Sibgrapi 2006) 6/23 Outline Problem description Motivation Local approximation From local to global approximation Results Future Works 7/23 Local approximations ? on the whole region 8/23 Local approximations Classical least squares method: Ax = B 9/23 Acceleration fitting We can improve the vector field approximation … 10/23 Acceleration fitting ... changing the minimization problem: µ user parameter 11/23 Robustness Least squares numerical instability Ridge regression: get around zero eigenvalues (A + βId)x = B β user parameter 12/23 Effects of Ridge Regression Without Ridge Regression With Ridge Regression 13/23 From local to global approximation 14/23 From local to global approximation Domain subdivision 15/23 Compact supports From local to global approximation Partition of Unity: and Kernel functions: 16/23 Results Synthetic field: 17/23 Results Stable Fluids: (Stam, J. 1999. Stable fluids) 18/23 Results Particle Image Velocimetry: (Luiz Fernando A. Azevedo, PUC-Rio) 19/23 Results Smoothed Particle Hydrodynamics (SPH): (F. Petronetto, A. Paiva, T. Lewiner, G. Tavares – PUC-Rio) 20/23 Conclusion & Future Work • Fast and accurate reconstruction → Extend to time dependent 2D vector fields • Applicable to visualization → Extend to time dependent 3D vector fields • Minimize quadratic error → More constraints: conservative, divergent-free vector fields reconstructions. 21/23 More results … The Method has already been extended to 3D vector fields 22/23 Thanks !!! 23/23