Filament Enhancement by Non-linear Volumetric Filtering Using Clustering-Based Connectivity
Shape filters are a family of connected morphological operators that have been used for filament enhancement in biomedical imaging. They interact with connected image regions rather than individual pixels, which can either be removed or retained unmodified. This prevents edge distortion and noise amplification, a property particularly appreciated in filtering and segmentation. In this paper we investigate their performance using a generalized notion of connectivity that is referred to as ”clustering-based connectivity”. We show that we can capture thin fragmented structures which are filtered out with existing techniques.
KeywordsHuman Head Difference Volume Mathematical Morphology Strong Cluster Magnetic Resonance Angiogram
Unable to display preview. Download preview PDF.
- 3.Orkisz, M., Hernández-Hoyos, M., Douek, P., Magnin, I.: Advances of blood vessel morphology analysis in 3D magnetic resonance images. Mach. Vis. Graph. 9, 463–471 (2000)Google Scholar
- 4.Ouzounis, G.K., Wilkinson, M.H.F.: Mask-based second-generation connectivity and attribute filters. In: IEEE Trans. Pattern Anal. Mach. Intell. (submitted, 2005)Google Scholar
- 8.Serra, J. (ed.): Image Analysis and Mathematical Morphology. II: Theoretical Advances. Academic Press, London (1988)Google Scholar
- 10.Urbach, E.R., Wilkinson, M.H.F.: Shape-only granulometries and grey-scale shape filters. In: Talbot, H., Beare, R. (eds.) Mathematical Morphology; Proc. 6th Int. Symp. Math. Morphology, Collingwood, Australia, pp. 305–314. CSIRO Publishing (2002)Google Scholar