An adaptive Kalman window filter to restore degraded images

摘要: A semicausal model for image representation has been described which accounts the correlated nature of pixel data. The is then used to develop a linear imaging system suitable Kalman algorithms. Since blurring PSF not known in practice, modified include estimation pixels while noise characteristics are assumed be known. For restoration, an adaptive filter developed whose length state vector shown function size resulting significant savings computational and storage requirements. Through examples, it demonstrated that by carefully choosing initial estimates error covariance terms, results comparable case when fully can obtained. Two criteria select such have described; one based on priori knowledge about dominant coefficient other law conservation light flux.