MULTISCALE MEDIAN AND MORPHOLOGICAL FILTERS FOR 2D PATTERN-RECOGNITION

Experiments demonstrate a multiscale decomposition that complements th ose using standard linear functions. It binds edges rather than waves to features of different scales. The configuration of non-linear media n or alternating sequential filters, 'morphological filters', used for the decomposition is referred to as a 'sieve'. Results suggest that w hilst some sieves produce an invertible transform, others have better statistical behaviour. Sieves are appropriate for isolating and locati ng the position of objects with sharp edges arising from nonlinear eve nts such as occlusion. They represent shape information in a way that is independent of spatial frequency, that has different uncertainty tr ade-offs, and can be used for signal analysis and pattern recognition. For example, by matching the granularity of an image with the granula rity of a target pattern, a simple pattern selective system (matched s ieve) can be implemented that outperforms its linear analogue, a match er filter. A sieve is a good multiscale smoother that improves upon si ngle step standard median and morphological filters and is particularl y appropriate for discontinuous signals, such as images where edges mu st be preserved.