Estimation and identification for 2-D block Kalman

The author considers the development of recursive identification and estimation techniques for the 2-D block Kalman filtering method proposed by M.R. Azimi-Sadjadi and P.W. Wong (IEEE Trans. on Acoustics, Speech, and Signal Proc., vol.ASSP-35, p.1736-49, Dec. 1987). A recursive identification scheme is introduced which can be used online to update image model parameters based upon the local statistics within a processing window. The correlation matrix of the model noise is estimated by solving a set of vector Yule-Walker equations. A recursive procedure is given for computing the parameters of the higher-order models. A method for updating the Kalman gain matrix which does not require any matrix inversion operation is also suggested.<>