Web Image Search by Automatic Image Annotation and Translation

There has been a growing interest in implementing online Web image search engine in the semantic level. However, most existing practical systems including popular commercial Web image search engines like Google and Yahoo! are either textbased or a simple hybrid of texts and visual features. This paper proposes a novel Web Image Search by Automatic Image Annotation and Translation (WISAIAT) system by using automatic image annotation and translation. We develop a technology which learns semantic image concepts from image contents and translates unstructured images into textual documents, so that images are indexed and retrieved in the same way as textual documents. Existing database management systems can be used to effectively manage image contents and image search can act as efficient as text search by translating images to textual documents through large scale machine learning. Experiments in both the Corel dataset and real Web dataset are performed to validate our system and the results are promising. This system suggests a new combination of texts and visual features to achieve a semantic Web image search and expected to become a reranking system to the existing Web image search result available online via the Internet. Keywordssemantic image search; automatic image annotation; image translation; Web image reranking; decision tree.