Recherche approximative de plus proches voisins

Sid-Ahmed Berrani 1 Laurent Amsaleg 2 Patrick Gros 2
2 TEXMEX - Multimedia content-based indexing
IRISA - Institut de Recherche en Informatique et Systèmes Aléatoires, Inria Rennes – Bretagne Atlantique
Abstract : Content-based retrieval is inherently an expensive process. It is possible, however, to reduce the cost of this process by searching for the approximate neighbors of the query points instead of searching for the exact result. This paper describes a new approach for performing efficient approximate nearest-neighbor searches in high-dimensional databases. It allows a fine and intuitive control over the precision of the search by setting the maximum probability to miss one of the exact nearest neighbors. In addition, we show that our approach is particularly well suited for image recognition based on local descriptors. The imprecision of individual nearest-neighbor searches is totally compensated by the multiplicity of queries.
Document type :
Journal articles
Complete list of metadatas

https://hal.inria.fr/inria-00604073
Contributor : Patrick Gros <>
Submitted on : Tuesday, June 28, 2011 - 9:59:06 AM
Last modification on : Friday, November 16, 2018 - 1:22:19 AM

Links full text

Identifiers

Citation

Sid-Ahmed Berrani, Laurent Amsaleg, Patrick Gros. Recherche approximative de plus proches voisins. Revue des Sciences et Technologies de l'Information - Série TSI : Technique et Science Informatiques, Lavoisier, 2003, 22 (9), pp.1201-1230. ⟨10.3166/tsi.22.1201-1230⟩. ⟨inria-00604073⟩

Share

Metrics

Record views

307