A recursive Kalman window approach to image restoration

Recursive restoration of blurred and noisy images using Kalman algorithms is hindered by the correlated nature of the two-dimensional data and the excessive computing requirements of the very long state vectors. This paper uses a semicausal model for image representation to account for the correlated nature of the data. Such a model is subsequently used to develop a discrete linear imaging system model, suitable for applying Kalman algorithms, to process the images in strips whose width is a function of the point spread function. This method of restoration results in considerable savings in computing time and storage. Two examples are presented to illustrate the technique.