PRELIMINARY AND SUBSEQUENT FILTERING OF NOISE IN IMAGE RESTORATION ALGORITHMS

The paper deals with noising of smeared or/and defocused images. The sequence of the noise filtering on such images is stated before elimination of the smearing/defocusing or after it. The concepts of preliminary and subsequent filtering of noise are brought in. The smearing/defocusing of a series of images is eliminated by the methods of parametric Wiener filtering and Tikhonov regularization, while the noise is filtered out by the methods of median Tuckey filtering and adaptive Wiener filtering. These questions are considered by carrying out experimental testing with obtaining numerical estimates of image restoration errors for different noise types and sequence of their filtering on representative samples. The methods of parametric Wiener filtering and Tikhonov regularization are shown to eliminate the smearing/defocusing effectively enough but noise filtering is not effective. Effectiveness of noise filtering is raised by adding such methods as the median Tuckey filter, the adaptive Wiener filter. Furthermore, an order (sequence) of filtering is important for impulse noise (before or after elimination of the smearing/defocusing depending on noise-signal situation), and for Gaussian noise the order is unessential. It is shown that an impulse noise is filtered better by the median, order, adaptive median filtering, and the Gaussian noise - by adaptive Wiener filtering, average filtering. An explanation for these effects is given.