Analysis of optical flow models in the framework of the calculus of variations

In image sequence analysis, variational optical flow computations require the solution of a parameter dependent optimization problem with a data term and a regularizer. In this paper we study existence and uniqueness of the optimizers. Our studies rely on quasiconvex functionals on the spaces W(1,p)(Omega, R(d)), with p > 1, BV(Omega, R(d)), BD(Omega). The methods that are covered by our results include several existing techniques. Experiments are presented that illustrate the behavior of these approaches.