Comparison of three filters in the solution of the Navier-Stokes equation in registration | Claus Gramkow, Morten Bro-Nielsen
| | Abstract | he registration of images is an often encountered problem in e.g. medical imagi
ng, satelite imaging, or stereo vision. In most applications a rigid deformation model does not suffce and complex deformations must be estimated. In medical registration Bajcsy et al. [1] have proposed the use of elastic models to describe the registration. Christensen et al. [3, 4, 5, 6, 8] proposed the use of fluid model
s that lack some of the constraints of the elastic model. They solve the viscous partial differential equation (PDE), which is the core problem of the fluid model, using the computationally expensive Succesive Over-Relaxation (SOR) algorithm. Thirion [9] calculated a flow velocity by regularizing the derived driving forces by a gaussian convolution filter. In this paper we propose an elastic filter to be used in convolution to approximate the solution of Navier-Stokes equation, and we compare the performance of the derived filter with two other filters, a separable approximation to the elastic filter and the well-known gaussian proposed by Thirion. The convolution approach is several times faster than the SOR algorithm. | | Keywords | registration, Navier-Stokes equation, convolution, elasticity filter. | | Type | Conference paper [With referee] | | Conference | Proc. Scandinavian Conference on Image Analysis (SCIA'97) | | Editors | Michael Frydrych and Jussi Parkkinen and Ari Visa | | Year | 1997 pp. 785-802 | | Address | Lappeenranta, Finland | | Electronic version(s) | [pdf] | | BibTeX data | [bibtex] | | IMM Group(s) | Image Analysis & Computer Graphics |
|