Extending area morphology to multivariate images
Adrian N. Evans · 2003
Area morphology has proved a popular technique for image filtering and segmentation.This paper investigates its extension to multivariate images.By defining the extrema of a connected set of vectors as the vector that is furthest from all other vectors, measured using a norm, regional vector extrema can be identified.These extrema are then processed in a manner analogous to scalar area morphology.This is achieved by the incorporation of additional constraints to ensure that each flat zone is treated as a single entity and to process the additional extrema created.Results demonstrate the effectiveness of this approach for filtering motion fields and colour images.