A combined Fuzzy and Probabilistic data descriptor for distributed CBIR

With the wide diffusion of digital image acquisition devices, the cost of managing hundreds of digital images is quickly increasing. Currently, the main way to search digital image libraries is by keywords given by the user. However, users usually add ambiguos keywords for large set of images. A content-based system intended to automatically find a query image, or similar images, within the whole collection is needed. In our work we address the scenario where medical image collections, which nowadays are rapidly expanding in quantity and heterogeneity, are shared in a distributed system to support diagnostic and preventive medicine. Our goal is to produce an efficient content-based description of each image collection in order to perform content-based image retrieval (CBIR) just in the node where the searched images are supposed to be. A novel combined fuzzy and probabilistic data descriptor is presented and experimental results are illustrated.

• Type: Conference
• Year: 2009
• Where: WILF
• Authors: Marco La Cascia, Marco Morana, Roberto Gallea
• Keywords: Distributed CBIR, Fuzzy Clustering, Medical Images