Recuperação por conteudo em grandes coleções de imagens heterogeneas

AUTOR(ES)
DATA DE PUBLICAÇÃO

2002

RESUMO

Content-based image retrieval (CBIR) is an area that has received increasing attention from the scientific community due to the exponential growing of available images, mainly at the WWW.This has spurred great interest for systems that are able to efficiently retrieve images according to their visual content. Our work has focused in techniques suitable for broad image domains. ln a broad image domain, it is not possible to assume or use any a p1 ior i knowledge about the visual content and/or semantic content of the images. Moreover, the cost of using semialitomatic image analysis techniques is prohibitive because of the heterogeneity and the amount of images that must be analyzed. We have directed our work to color-based image retrieval, and have focused on the three main issues that should be addressed in order to achieve color-based image retrieval: (1) how to analyze and describe images in an automatic and efficient way; (2) how to represent the image content in a compact and effective way; and (3) how to efficiently compare the visual features extracted from the images. The main contributions of our work are two algorithms to automatically analyze the visual content of the images (CBC and BIC), two distance functions to compare the visual features extracted from the images (MiCRoM and dLog), and an alteruative representation for CBIR approaches that decompose and represent images according to a grid of equalsized cells (CCH)

ASSUNTO(S)

processamento de imagens recuperação da informação banco de dados

ACESSO AO ARTIGO

Documentos Relacionados