6533b7d3fe1ef96bd125fe86
RESEARCH PRODUCT
An unsupervised region growing method for 3D image segmentation
Vito Di GesùFranco Chiavettasubject
business.industryRegion growingComputer sciencePartition (number theory)Fuzzy conceptPattern recognitionArtificial intelligenceExtreme pointbusinessReal imageDecomposition problemdescription
The paper deals with 3D shape decomposition problem, objects are modelled as finite unions of almost-convex primitives. A new region growing method is proposed to extract meaningful objects parts. Parts are individuated by performing a set-partitioning of surface dominating points. The partition step returns labelled seeds from which to start a region growing procedure that propagate labels onto object surface patches. A fuzzy concept of λ-convexity is introduced to test noised real images. Experimental results are given.
| year | journal | country | edition | language |
|---|---|---|---|---|
| 1995-01-01 |