Dual cross-media relevance model for image annotation

• DOI: 10.1145/1291233.1291380
• Publication Date: 2007-10-14T12:51:38Z
• Authors: Jing Liu; Bin Wang; Mingjing Li; Zhiwei Li; Weiying Ma; Hanqing Lu; Songde Ma
• Keywords: Relevance

Image annotation has been an active research topic in recent years due to its potential impact on both image understanding and web retrieval. Existing relevance-model-based methods perform by maximizing the joint probability of images words, which is calculated expectation over training images. However, semantic gap dependence data restrict their performance scalability. In this paper, a dual cross-media relevance model (DCMRM) proposed for automatic annotation, estimates words pre-defined lexicon. DCMRM involves two kinds critical relations annotation. One word-to-image relation other word-to-word relation. Both can be estimated using search techniques as well available data. Experiments conducted Corel dataset demonstrate effectiveness model.