Boundary element methods for solving Poisson equations in computer vision problems
Gendai Gu, Michael A. Gennert · 2002
The boundary element method (BEM) for solving Poisson's equations is described. Issues in BEM, such as Green's functions, boundary conditions, evaluation of improper integrals, and continuity up to first derivative of solution functions, are discussed. BEM is compared with FEM, the finite element method, in terms of storage and time complexity. The authors discuss application to vision: height from gradient; shape from shading; surface interpolation; brightness based stereo matching; and the optical flow problem. Brief mention is made of some early experimental results on synthetic images.>