A new linear model for image representation for use with Kalman filter restoration

When a Kalman filter is applied to an image for restoration purposes, the model of the original image affects the accuracy of the restoration. The model for effective restoration depends on the correlation of the original image and the variance of the noise. If these parameters are unknown, they have to be estimated from the observed image. In this paper, a method to estimate the unknown parameters in the image restoration process is proposed along with a method that identifies the region of support (number of pixels used for estimation and their positions). Three schemes were developed and used to represent the image and to identify the parameters for the filter. The approach has been tested on numerous images. Results show superior performance compared to methods and implementations previously reported in the literature both in terms of computational complexity and signal to noise ratio.