A carregar...
Publicação
Recuperação de imagens digitais com base no conteúdo: estudo na Biblioteca de Arte e Arquivos da Fundação Calouste Gulbenkian
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 6.49 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
O crescimento massivo de dados multimédia na Internet e surgimento de novas
plataformas de partilha criou grandes desafios para a recuperação de informação. As
limitações de pesquisas com base em texto para este tipo de conteúdo proporcionaram
o desenvolvimento de uma abordagem de recuperação de informação com base no
conteúdo (CBIR) que recebeu atenção crescente nas últimas décadas.
Tendo em conta as pesquisas realizadas nesta área, e sendo o foco desta investigação
as imagens digitais, são explorados conceitos e técnicas associadas a esta abordagem
por meio de um levantamento teórico que relata a evolução da recuperação de
informação e a importância que esta temática tem para a Gestão e Curadoria da
Informação. No contexto dos sistemas que têm vindo a ser desenvolvidos recorrendo
à indexação automática, indicam-se as diversas aplicações deste tipo de processo.
São também identificadas ferramentas relacionadas disponíveis para a realização de
um estudo da aplicação deste tipo de recuperação de imagens no contexto da
Biblioteca de Arte e Arquivos da Fundação Calouste Gulbenkian e as coleções
fotográficas que esta é detentora no seu acervo, considerando as particularidades da
instituição em que estas se inserem. Para a demonstração pretendida e de acordo com
os critérios estabelecidos, recorreu-se inicialmente a soluções CBIR disponíveis em
linha e, numa fase seguinte, foi usada uma ferramenta de instalação local numa
coleção específica.
Através deste estudo, são atestados os pontos fortes e pontos fracos da recuperação de
imagens digitais com base no conteúdo face à abordagem mais tradicional com base
em metainformação textual em vigor atualmente nessas coleções. Tendo em
consideração as necessidades dos utilizadores dos sistemas em que estes objetos
digitais se encontram indexados, a combinação entre estas técnicas pode originar
resultados mais satisfatórios.
The massive growth of multimedia data on the Internet and the emergence of new sharing platforms created major challenges for information retrieval. The limitations of text-based searches for this type of content have led to the development of a content-based information retrieval approach that has received increasing attention in recent decades. Taking into account the research carried out in this area, and digital images being the focus of this research, concepts and techniques associated with this approach are explored through a theoretical survey that reports the evolution of information retrieval and the importance that this subject has for Information Management and Curation. In the context of the systems that have been developed using automatic indexing, the various applications of this type of process are indicated. Available CBIR tools are also identified for a case study of the application of this type of image retrieval in the context of the Art Library and Archives of the Calouste Gulbenkian Foundation and the photographic collections that it holds in its resources, considering the particularities of the institution to which they belong. For the intended demonstration and according to the established criteria, online CBIR tools were initially used and, in the following phase, locally installed software was selected to search and retrieve in a specific collection. Through this case study, the strengths and weaknesses of content-based image retrieval are attested against the more traditional approach based on textual metadata currently in use in these collections. Taking into consideration the needs of users of the systems in which these digital objects are indexed, combining these techniques may lead to more satisfactory results.
The massive growth of multimedia data on the Internet and the emergence of new sharing platforms created major challenges for information retrieval. The limitations of text-based searches for this type of content have led to the development of a content-based information retrieval approach that has received increasing attention in recent decades. Taking into account the research carried out in this area, and digital images being the focus of this research, concepts and techniques associated with this approach are explored through a theoretical survey that reports the evolution of information retrieval and the importance that this subject has for Information Management and Curation. In the context of the systems that have been developed using automatic indexing, the various applications of this type of process are indicated. Available CBIR tools are also identified for a case study of the application of this type of image retrieval in the context of the Art Library and Archives of the Calouste Gulbenkian Foundation and the photographic collections that it holds in its resources, considering the particularities of the institution to which they belong. For the intended demonstration and according to the established criteria, online CBIR tools were initially used and, in the following phase, locally installed software was selected to search and retrieve in a specific collection. Through this case study, the strengths and weaknesses of content-based image retrieval are attested against the more traditional approach based on textual metadata currently in use in these collections. Taking into consideration the needs of users of the systems in which these digital objects are indexed, combining these techniques may lead to more satisfactory results.
Descrição
Palavras-chave
Recuperação de imagens Recuperação de informação Content-based image retrieval Metainformação Text-based image retrieval Image retrieval Information retrieval Automatic indexing