Retrieval of multimedia documents using low cost image ranking algorithms

Huge collections of digital multimedia documents are available to computer users through the global Internet and cheap high volume storage and distribution media. Manual indexing of these massive collections is prohibitively expensive, especially when a collection is dynamic; automatic indexing is essential to support multimedia retrieval. Text and images are the most common components of digital documents, but existing search engines do not utilize image evidence in their ranking algorithms. In this paper, we investigate the ranking of digital documents using image features. Experimental results show that multimedia retrieval using image-ranking algorithms can provide higher effectiveness than the retrieval effectiveness of algorithms based on text ranking algorithms for a given type of information needs.