Skip to main content

Selection of Segment Similarity Measures for Hierarchical Picture Segmentation

  • Conference paper
Book cover Computer-Generated Images
  • 147 Accesses

Résumé

Différentes mesures de la similarité entre segments sont étudiées pour la segmentation de l’image. Une segmentation agglomerative hiérarchique compare et fusionne deux segments s’ils sont similaires. L’algorithme utilisé fusionne, à chaque itération, les deux segments les plus similaires. Regardant la segmentation de l’image comme un problème d’approximation, la mesure de similarité est reliée à l’erreur totale d’approximation; l’accroissement de l’erreur produit par la fusion de deux segments est utilisé. Une image de télédétection est segmentée par l’approximation de chaque région soit par une constante ou soit par un plan. Une mesure de similarité sensible à la variance locale de l’image est aussi présentée. Les avantages de la combinaison de différentes mesures sont aussi soulignés. Différentes régions d’une image peuvent nécessiter différentes mesures qui doivent donc être combinées pour obtenir, dans l’ensemble, un bon résultat. De plus, dans une segmentation hiérarchique, une mesure simple peut être utilisée lors des premières fusions, tandis que des mesures plus complexes vont être employees par la suite.

Abstract

The problem of defining appropriate segment similarity measures for picture segmentation is exmained. In agglomerative hierqarchical segmentation, two segments are coamapared and merged if found similar. The propesed Hierarchical Step-Wise Optimization (HSWO) algorithm finds and then merges the two most similar segements, on a step-by-step basis. By considering picture segmentation as a piece-wise picture approximation problem, the similarity measure (or the step-wise criterion) is related to the overall approximation error. The measure then corresponds to the increase of the approximation error resulting from merging two segments. Similarity measures derived from constant approximations (zeroth order polynomials) and planar approximations (first order polynomials). An adaptive measurebased upon local variance is also used. The advantages of combining similarity measures (or cirteria) are also stressed. Different picture areas can require different measures which must therefore be combined in order to obtain good overall results. Moreover, in hierarchical segmentation, simple measures can be used for the first merging steps, while, at a higher level of the segment hierarchy, more complex measures can be employed.

This is a preview of subscription content, log in via an institution to check access.

Access this chapter

Chapter
USD 29.95
Price excludes VAT (USA)
  • Available as PDF
  • Read on any device
  • Instant download
  • Own it forever
eBook
USD 39.99
Price excludes VAT (USA)
  • Available as PDF
  • Read on any device
  • Instant download
  • Own it forever
Softcover Book
USD 54.99
Price excludes VAT (USA)
  • Compact, lightweight edition
  • Dispatched in 3 to 5 business days
  • Free shipping worldwide - see info

Tax calculation will be finalised at checkout

Purchases are for personal use only

Institutional subscriptions

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  • Beaulieu JM, Goldberg M (1982) Hierarchical Picture Segmentation by Approximation. Canadian Communication and Enttigy Convenience, Montreal (Canada): 393–396.

    Google Scholar 

  • Beaulieu JM, Goldberg M (1983) Step-Wise Optimization for Hierarchical Picture Segmentation. Conf. on ComputeVision and atte.fin Recognition, Washington, D.C.: 59–64.

    Google Scholar 

  • Brice C, Fennema C (1970) Scene Analysis Using Regions. A Intell. 1: 205–226.

    Article  Google Scholar 

  • Chen PC, Pavlidis T (1980) Image Segmentation as an Estimation Problem. Comput. Graphic image, Pioce,66. 12: 153–172.

    Article  Google Scholar 

  • Freuder EC (1976) Affinity: A Relative Approach to Region Finding. Comput. Griaphics Image. Piece. 65: 254–264.

    Article  Google Scholar 

  • Horowitz SL, Pavlidis T (1976) Picture Segmentation by a Tree Traversal Algorithm. JACM 23: 368–388.

    Article  MATH  Google Scholar 

  • Pavlidis T (1979) Hierarchies in Structural Pattern Recognition. Koc. IEEE 67:737–744.

    Google Scholar 

  • Zobrist AL, Thompson WB (1975) Building a Distance Function for Gestalt Grouping. IEEE trans. Comput C-4: 718–728.

    Article  Google Scholar 

Download references

Author information

Authors and Affiliations

Authors

Editor information

Editors and Affiliations

Rights and permissions

Reprints and permissions

Copyright information

© 1985 Springer-Verlag Tokyo

About this paper

Cite this paper

Beaulieu, J.M., Goldberg, M. (1985). Selection of Segment Similarity Measures for Hierarchical Picture Segmentation. In: Magnenat-Thalmann, N., Thalmann, D. (eds) Computer-Generated Images. Springer, Tokyo. https://doi.org/10.1007/978-4-431-68033-8_8

Download citation

  • DOI: https://doi.org/10.1007/978-4-431-68033-8_8

  • Publisher Name: Springer, Tokyo

  • Print ISBN: 978-4-431-68035-2

  • Online ISBN: 978-4-431-68033-8

  • eBook Packages: Springer Book Archive

Publish with us

Policies and ethics