GC-MORSE: Generalized Connected Morphological Operators for Robust Shape Extraction

Summary

The aim of this project is to explore multi-scale and shape-based morphological connected operators, in terms of theory, algorithm development and robustness in a real application, with the main emphasis on the latter two. Connected filters allow image filtering without distorting edges, a feature which is highly desirable in many applications. For this reason they have received much attention in the literature in recent years. Usually, connected filters are defined by classical 4- or 8-connectivity (in 2-D). However, several theoretical developments concerning generalizations of the notion of connectivity have been proposed, which might improve the robustness and increase the versatility of these filters. In this project we will explore the relationships between these various proposals, and aim to develop efficient, unified algorithms for computing connected filters based on a variety of generalized connectivities. After this, algorithms for obtaining shape-scale and orientation-scale spaces using connected filters will be developed. In both cases we will build on a number of algorithms developed here already. These new algorithms will be applied to the problem of enhancing details in 2-D and 3-D medical images, in particular the detection of filamentous details such as blood vessels simultaneously at multiple scales.

Composition of Research Team

Dr. M. H. F. Wilkinson Principal investigator and referent Institute for Mathematics and Computing Science, University of Groningen.
Prof. Dr. N. Petkov Promotor Institute for Mathematics and Computing Science, University of Groningen.
G.K.Ouzounis PhD student Institute for Mathematics and Computing Science, University of Groningen.

Publications

  1. G.K. Ouzounis and M.H.F. Wilkinson. Hyperconnected attribute filters based on k-flat zones for 3D medical imaging. IEEE Trans. Pattern Anal. Mach. Intell. submitted.

  2. G.K. Ouzounis and M.H.F. Wilkinson. Partition-induced connections and operators for pattern analysis. Pattern Recognition, submitted.

  3. G.K. Ouzounis and M.H.F. Wilkinson. A parallel dual-input max-tree algorithm for shared memory machines. Proc. Int. Symp. Math. Morph. (ISMM) 2007, Oct 10-13, 2007, Rio de Janeiro, pp. 449-460. PDF version (2.0 MB) available.

  4. G.K. Ouzounis and M.H.F. Wilkinson. Mask-Based Second Generation Connectivity and Attribute Filters. IEEE Trans. Pattern Anal. Mach. Intell. (2007) 29(2): 990-1004. PDF preprint version (844 kB) available.

  5. G.K. Ouzounis and M.H.F. Wilkinson. Filament enhancement by non-linear volumetric filtering using clustering-based connectivity. Lecture Notes in Computer Science Vol. 4153 International Workshop on Intelligent Computing in Pattern Analysis/Synthesis, IWICPAS 2006, (N. Zheng, X. Jiang, and X. Lan, Eds.), Xi'an, China, August 2006, pp. 317-327.

  6. G.K. Ouzounis and M.H.F. Wilkinson. Countering Oversegmentation in Partitioning-Based Connectivities. Int. Conf. Image Proc. 2005 Vol. III , September 11-14, Genova, Italy, pp.844-847. PDF-version (283 kB) available.

  7. G.K. Ouzounis and M.H.F. Wilkinson. Second-Order Connected Attribute Filters Using Max-Trees. Mathematical Morphology: 40 years on, Proc. Int. Symp. Math. Morph. (ISMM) 2005, pp. 65-74.

Last modified: Wed Jan 14 09:59:03 CET 2004