Gradient filters based on the fast wavelet transform for quasi-identical noisy images

The paper deals with a family of nonlinear gradient filters that can be applied to one noisy image or several quasi-identical patterns (i.e., several images of the same object with independent noise). Such filters are based on a specific wavelet decomposition and a preliminary statistical analysis of quasi-identical patterns. Results of numerical experiments have shown that the developed gradient filters reduce noise and preserve image boundaries.