Multiscale decomposition using median and morphological filters
Jenny Bangham, Richard V. Aldridge · 2005
The small differences between the output from cascades of multiscale median and morphological filters (datasieves) is explored. It is shown that the self dual nature median filters confers statistical verity and that open/close (close/open) filters have the advantage of yielding what appears to be a 1:1 transform pair. Overall, however, they distinguish ID patterns equally well from an uncorrelated noisy background.