Book chapter
Electric Field Theory Motivated Graph Construction for Optimal Medical Image Segmentation
Graph-Based Representations in Pattern Recognition, pp.334-342
Lecture Notes in Computer Science, Springer Berlin Heidelberg
2009
DOI: 10.1007/978-3-642-02124-4_34
Abstract
In this paper, we present a novel graph construction method and demonstrate its usage in a broad range of applications starting from a relatively simple single-surface segmentation and ranging to very complex multi-surface multi-object graph based image segmentation. Inspired by the properties of electric field direction lines, the proposed method for graph construction is inherently applicable to n-D problems. In general, the electric field direction lines are used for graph “column” construction. As such, our method is robust with respect to the initial surface shape and the graph structure is easy to compute. When applied to cross-surface mapping, our approach can generate one-to-one and every-to-every vertex correspondent pairs between the regions of mutual interaction, which is a substantially better solution compared with other surface mapping techniques currently used for multi-object graph-based image segmentation.
Details
- Title: Subtitle
- Electric Field Theory Motivated Graph Construction for Optimal Medical Image Segmentation
- Creators
- Yin Yin - University of IowaQi Song - University of IowaMilan Sonka - University of Iowa
- Resource Type
- Book chapter
- Publication Details
- Graph-Based Representations in Pattern Recognition, pp.334-342
- Series
- Lecture Notes in Computer Science
- DOI
- 10.1007/978-3-642-02124-4_34
- eISSN
- 1611-3349
- ISSN
- 0302-9743
- Publisher
- Springer Berlin Heidelberg; Berlin, Heidelberg
- Language
- English
- Date published
- 2009
- Academic Unit
- Roy J. Carver Department of Biomedical Engineering; Electrical and Computer Engineering; Radiation Oncology; Fraternal Order of Eagles Diabetes Research Center; Injury Prevention Research Center; Ophthalmology and Visual Sciences
- Record Identifier
- 9984186706602771
Metrics
16 Record Views