Boundary value selection problem for image restoration using the reduced order model based Kalman filter

The reduced order-model Kalman filter (ROMKF) is a low order state-space model based Kalman filter. The motivation for introducing the ROM was the reduction in the amount of computation involved in a 2-D Kalman filter with full state-space model representation. Because of the way in which the state vector and the covariance are defined in the ROM, it is necessary to give careful consideration to the selection of the 2-D boundary conditions. A discussion is presented of such considerations, and it is shown, using both error indices and visual results, that proper boundary selection will significantly improve image restoration