Ringing reduction in images restoration using mirror images and adaptive Kalman filtering

摘要

Recursive 2D Kalman filtering has been successfully used in image restoration. In order to reduce computational work and large amounts of storage, a fast algorithm has been introduced by Biemond. However this method is based on the DFT transform and ringing on the restored image results from boundary conditions. Furthermore, the Kalman gain factors may increase due to the zeros of the Fourier transform of the PSF periodicity is realized by means of a new image mirror. No more computational work is required since this new image is symmetric. Then the authors introduce adaptive Kalman filtering in order to avoid divergence of Kalman gain factors and to reduce the remaining ringing.