Comparison and evaluation of a multi-tag method for advanced image search

Abstract

Tagging is widely applied to web documents, such as blogs, and multimedia data, such as images and video; however, insufficient results appear during image retrieval using tags because of the fundamental limitations of tags. In this paper, a multi-tag based search algorithm is proposed to overcome the above weakness. The proposed algorithm performs image search using a 2-tag, 3-tag and 3-2 tag combined search methods. These multi-tag search algorithms are compared and evaluated with the earlier method from the point of view of correctness and precision. Searched images from flickr, a typical tagging site, are used for this examination. The results of the examination show that the correctness of multi-tag based image search is higher than the existing single-tag search methods. The precision and recall of existing single-tag search methods is 48.7%, while the precision and recall of multi-tag algorithms varies by method; however, the average is 83.6% in precision and 48.7% in recall. © 2017 International Information Institute.