A nonlinear multiscale pyramidal transform based on nonoverlapping block de
compositions using the median operation and a polynomial approximation is c
onsidered. It is shown that this structure can be useful for denoising of o
ne- and two-dimensional (1-D and 2-D) signals, Various denoising techniques
are analyzed, including methods based on spatially adaptive thresholding a
nd partial cycle-spinning algorithms. An analytical method for deriving the
distribution function of the transform coefficients is also presented, Thi
s, in turn, can be used for the selection of thresholds for denoising appli
cations.