A new procedure, where the original image and blur parameters can be simultaneously identified using two-dimensional (2-D) hybrid lattice structures has been presented, utilizing the fact that noisy and blurred images be presented by an autoregressive moving average (ARMA) model. AR parameters obtained as the result of identification process, represent the original image, whereas the MA parameters represent the blur. In this proposed method, the noncausal blur has been processed in the support region covering the whole plane, without the need for factorization into quarter planes. Besides, there is no need for normalization regarding the blur parameters. Since the method is recursive in nature, it is not affected by image data size. It is expected that the obtained image and blur parameters can be used for image restoration purposes. © 2005 IEEE.