In this paper we propose a method to detect abrupt shot changes in MPEG coded videos that operates directly on the compressed domain by using a Multi-Expert approach. Generally, costly analysis for addressing the weakness of a single expert for abrupt shot change detection and the consequent modifications would produce only slight performance improvements. Hence, after a careful analysis of the scientific literature, we selected three techniques for cut detection, which extract complementary features and operate directly in the compressed domain. Then, we combined them into different kinds of Multi-Expert Systems (MES) employing three combination rules: Majority Voting, Weighted Voting and Bayesian rule. In order to assess the performance of the proposed MES, we built up a huge database, much wider than those used in the field. Experimental results demonstrate that the proposed system performs better than each of the three single algorithms.

Automatic Cut Detection in MPEG Movies: a Multi-Expert Approach

DE SANTO, Massimo;PERCANNELLA, Gennaro;VENTO, Mario
2002

Abstract

In this paper we propose a method to detect abrupt shot changes in MPEG coded videos that operates directly on the compressed domain by using a Multi-Expert approach. Generally, costly analysis for addressing the weakness of a single expert for abrupt shot change detection and the consequent modifications would produce only slight performance improvements. Hence, after a careful analysis of the scientific literature, we selected three techniques for cut detection, which extract complementary features and operate directly in the compressed domain. Then, we combined them into different kinds of Multi-Expert Systems (MES) employing three combination rules: Majority Voting, Weighted Voting and Bayesian rule. In order to assess the performance of the proposed MES, we built up a huge database, much wider than those used in the field. Experimental results demonstrate that the proposed system performs better than each of the three single algorithms.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

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: http://hdl.handle.net/11386/1128947
 Attenzione

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

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 2
  • ???jsp.display-item.citation.isi??? ND
social impact