Project FMI-SiR



The FMI-SiR (user-defined Features, Metrics and Indexes for Similarity Retrieval), is a DBMS module developed to perform data selection on a relation executing similarity search operations. It includes the required mechanisms to perform similarity search into a database core to efficiently execute similarity queries. The implementation of the FMI-SiR using the Oracle's Extensible Architecture Framework to handle complex data is called the FMI-SiRO and its extension to handle DICOM medical images is called Medical FMI-SiR (MedFMI-SiR).




The FMI-SiRO is attached to the DBMS using the Oracle's extensibility interfaces, providing similarity retrieval functionalities to the client applications in a transparent way. Client applications interact with the module accessing directly the DBMS using SQL, without requiring any additional software libraries. This module allows performing queries combining metadata- and content-based conditions, providing metrics and complex data indexes. It is controlled by the DBMS query processor, thus providing a tight integration to other DBMS operations. The index capabilities are achieved integrating the Arboretum library through the Oracle extensibility indexing interface, providing an efficient processing of similarity queries. Applications employing complex data can define its own domain specific feature extractors and/or distance functions, whereas using the powerful resources of the Oracle database in an integrated way.


The MedFMI-SiR extension interacts with the DICOM Management Library, which is the DICOM Toolkit (DCMTK) in the current implementation, which is a collection of C++ libraries and applications implementing the DICOM standard, including software to examine, construct and convert image files into/from the DICOM format. This module is capable of opening several modalities of DICOM images, including compressed images, to perform image processing operations and it can serve multiple applications, such as PACS, DICOM viewers and other already existing data analysis applications.








Contact Prof. Daniel S. Kaster ()