Geometric Verification (GV) is the last step for most visual search systems. It consists of two parts: first, ratio test is used to find matches between feature descriptors; second, a geometric consistency check is applied. Both steps are computationally expensive, but all the attempts made to speed up the process deal with the geometric check part only. In this work, we focus indeed on ratio test. Using simple PCA and other tricks, a speed-up of an order of magnitude is achieved preserving good retrieval accuracy. Moreover, we propose a modified ranking approach which exploits distance information between descriptors and further improves retrieval performance.

A New Ranking Approach and a Revisited Ratio Test for Improving Content-Based Image Retrieval

MAURO, Massimo;LEONARDI, Riccardo
2012-01-01

Abstract

Geometric Verification (GV) is the last step for most visual search systems. It consists of two parts: first, ratio test is used to find matches between feature descriptors; second, a geometric consistency check is applied. Both steps are computationally expensive, but all the attempts made to speed up the process deal with the geometric check part only. In this work, we focus indeed on ratio test. Using simple PCA and other tricks, a speed-up of an order of magnitude is achieved preserving good retrieval accuracy. Moreover, we propose a modified ranking approach which exploits distance information between descriptors and further improves retrieval performance.
2012
9781467325332
File in questo prodotto:
File Dimensione Formato  
ML_ICIP-2012_full-text.pdf

solo utenti autorizzati

Descrizione: ML_ICIP-2012_full-text
Tipologia: Full Text
Licenza: NON PUBBLICO - Accesso privato/ristretto
Dimensione 249.68 kB
Formato Adobe PDF
249.68 kB Adobe PDF   Visualizza/Apri   Richiedi una copia
ML_ICIP-2012_pre-print.pdf

accesso aperto

Descrizione: ML_ICIP-2012_pre-print
Tipologia: Documento in Pre-print
Licenza: Creative commons
Dimensione 345.4 kB
Formato Adobe PDF
345.4 kB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11379/164833
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? 0
social impact