Second Order Connected Atrribute Filters

Georgios Ouzounis
Intelligent Systems Group

Abstract:

The notion of connectivity in images characterizes the way pixels are grouped to form connected components or flat-zones in gray-scale. In mathematical morphology, connected operators have been developed which perform filtering based on these groupings. Connected operators modify the intensity of connected image regions instead of individual pixels. This prevents edge distortion, a property highly desirable in many applications. Second-Order connectivity is an extension to conventional connectivity providing a flexible approach in handling groups of connected sets as single clusters or object regions bridged by narrow elongated segments as individual sets. These two perceptual conditions referred to as clustering and partitioning respectively, are discussed in this talk with an extension to generalized attribute filters. The two conditions are summarized in a new theoretical framework developed within the GC-MORSE project which will be presented along with a demonstration of the potential of these operators

back to the list of talks