Multi Feature Indexing Network MUFIN for Similarity Search Applications

Similarity has been a central notion throughout our lives and due to the current unprecedented growth of digital data collections of various types in huge quantities, similarity management of digital data is becoming necessary. The Multi-Feature Indexing Network (MUFIN) is a generic engine for similarity search in various data collections, such as pictures, video, music, biornetric data, sensor and scientific data, and many others. MUFIN can provide answers to queries based on the example paradigm. The system assumes a very universal concept of similarity that is based on the mathematical notion of metric space. In this concept, the data collection is seen as objects together with a method to measure similarity between pairs of objects. The key implementation strategies of MUFIN concern: extensibility - to be applied on variety of data types, scalability - to be efficient even for very large databases, and infrastructure independence - to run on various hardware infrastructures so that the required query response time and query execution throughput can be adjusted. The capability of MUFIN is demonstrated by several applications and advance prototypes. Other applications and future research and application trends are also to be discussed.