Image Similarity Search: Theory and Practice
| Authors | |
|---|---|
| Year of publication | 2007 |
| Type | Article in Proceedings |
| Conference | MEMICS 2007: Third Doctoral Workshop on Mathematical and Engineering Methods in Computer Science |
| MU Faculty or unit | |
| Citation | |
| web | MEMICS |
| Field | Informatics |
| Keywords | similarity search; content based image retrieval; peer-to-peer |
| Description | The data-explosion phenomenon proceeds in two respects: (1) The volume of data produced is increasing rapidly and (2) new data types appear and are widely used. This calls for development of brand new indexing and searching methods which would respect the needs of the recent data types and be efficient on vast amounts of data. This paper describes a transfer of our previous theoretical results in this area into practice by building a fully functional application. The application is able to efficiently manage large collections of digital images and search these images according to their very content (the similarity search). Its distributed architecture is based on the peer-to-peer paradigm and the searching method adopts the metric-based approach to similarity. Currently the application can store and search tens of millions of images downloaded from the Web with dozens of simultaneous users, although it runs on a limited hardware infrastructure. |
| Related projects: |