Skip to Main Content
In this paper, a maximum likelihood estimation approach is presented for the identification and restoration of images degraded by noise and blur. By employing the unitary discrete sine transform (DST), the spatial state-space representation of the noisy blurred image is completely decoupled into a bank of one-dimensional real state-space subsystems, to be identified. Then, the maximum likelihood estimation technique using the expectation-maximization (EM) algorithm is developed to jointly identify the blurring functions, the image model parameters and the noise variance. The computational efficiency is implemented by incorporating the conventional Kalman smoothed estimators. The original image is restored by the inverse DST of the transformed image estimates. It is shown by the experimental results that the restoration effect due to the proposed method is superior to that by the recently proposed DFT based method.