Cut-Region: a compact building block for hierarchical metric indexing

  • Authors:
  • Jakub Lokoč;Přemysl Čech;Jiří Novák;Tomáš Skopal

  • Affiliations:
  • SIRET research group, Dept. of Software Engineering, Faculty of Mathematics and Physics, Charles University in Prague, Czech Republic;SIRET research group, Dept. of Software Engineering, Faculty of Mathematics and Physics, Charles University in Prague, Czech Republic;SIRET research group, Dept. of Software Engineering, Faculty of Mathematics and Physics, Charles University in Prague, Czech Republic;SIRET research group, Dept. of Software Engineering, Faculty of Mathematics and Physics, Charles University in Prague, Czech Republic

  • Venue:
  • SISAP'12 Proceedings of the 5th international conference on Similarity Search and Applications
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

With the emerging applications dealing with complex multimedia retrieval, such as the multimedia exploration, appropriate indexing structures need to be designed. A formalism for compact metric region description can significantly simplify the design of algorithms for such indexes, thus more complex and efficient metric indexes can be developed. In this paper, we introduce the cut-regions that are suitable for compact metric region description and we discuss their basic operations. To demonstrate the power of cut-regions, we redefine the PM-Tree using the cut-region formalism and, moreover, we use the formalism to describe our new improvements of the PM-Tree construction techniques. We have experimentally evaluated that the improved construction techniques lead to query performance originally obtained just using expensive construction techniques. Also in comparison with other metric and spatial access methods, the revisited PM-Tree proved its benefits.