A Novel Semantics-Based Image Retrieval Method Using Similarity Measure of Multi-Level Semantics
Shiliang Chen,Zhanhuai Li,Liu Yuan
DOI: https://doi.org/10.3969/j.issn.1000-2758.2008.05.012
2008-01-01
Xibei Gongye Daxue Xuebao/Journal of Northwestern Polytechnical University
Abstract:Aim. Words in image label frequently have multi-senses, leading to unsuitable search results for users. We now propose using WordNet to disambiguate image senses. In section 1 of the full paper, we discuss semantic hierarchy of images. Section 2 explains the calculation of image semantic similarity measure. What we do in section 2 is essentially that, we, employing the multi-relations between words, such as synonyms, hyponyms and hypernyms provided by WordNet, build the relationships between image labels, thus unifying the description of image label. The image semantic similarity measure can be calculated by eq. (7). In section 3, we give experimental results shown in Fig. 1 and Table 1, which indicate preliminarily that the calculated image semantic similarity measure in our semantics-based image retrieval method gives, in general, better performance as compared with that of the label-based method.