Guaatupi: um ambiente para indexação e recuperação de imagens da web sem redundância visual

• Thiago Fonsêca Meneses UFCG
• Carlos A. F. Pimentel Filho UFMG
• Rafael W. M. de Araujo US

Abstract

This work presents an approach for image retrieval system based on its visual features applied to World Wide Web scenario. The proposed approach presents techniques for crawling, indexing and retrieving without visual redundance image copies. In order to achieve its goals, the present work uses a spider to capture web images from Google Images and wavelets transform in order to extract the essence of image features that well represents them. In the present work, it was indexed 310 thousand images, which were reduced to 225 thousand after image copies elimination. It represents a reduction in about 27% of redundancy and 20% faster.