Name
Abstract | ||
|---|---|---|
Perceptual organization has two essential factors that affect the grouping result directly: how to extract grouping cues and how to grouping. In this paper, a global contour-grouping algorithm based on spectral clustering is presented. First, a new grouping cue called wavelet edge is obtained in multi-scale space, which not only has the property of intensity and direction, but also has the property of singularity measured by lipschitz exponent. Thus grouping cues carry the information of both areas and edges. Secondly, a global grouping approach is presented by use of spectral clustering that has no limitation of neighborhood. Furthermore, the Gestalt principles are used to optimize the grouping result by adding penalty item in iterative process. The experiments show that this algorithm will be effective on condition that the singularities of the edges that belong to one object are equal or close, especially for partially occluded object. |
Year | DOI | Venue |
|---|---|---|
2008 | 10.1007/978-3-540-87734-9_51 | ISNN (2) |
Keywords | Field | DocType |
iterative process,new grouping cue,spectral clustering,gestalt principle,occluded object,essential factor,global contour-grouping algorithm,grouping result,grouping cue,global grouping approach,scale space,singularity | Spectral clustering,Pattern recognition,Iterative and incremental development,Computer science,Algorithm,Singularity,Gestalt psychology,Lipschitz exponent,Artificial intelligence,Gravitational singularity,Perception,Wavelet | Conference |
Volume | ISSN | Citations |
5264 | 0302-9743 | 0 |
PageRank | References | Authors |
0.34 | 7 | 3 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Hui Yin | 1 | 0 | 0.68 |
| Siwei Luo | 2 | 244 | 41.85 |
| Yaping Huang | 3 | 108 | 21.45 |