Metadata projection for visual resources retrieval

Abstract

The amount of multimedia documents available in the net stimulates a strong need to explore metadata for improving efficiency of systems’ performance whether commercial, prototypical or open source. This paper provides a framework on image retrieval in cross-language environment based on metadata. Digital images provide visual resources information content that is known as content-based image retrieval (CBIR). We investigate the correlation between query linguistic characteristics that match the visual object collections and the image databases as well as some indexing strategies. Our test set forms a baseline to analyze influence of English and French phenomena on the strategy of browsing the Web for image collections in multilingual environment. Metadata in our experiment is defined as keywords, annotations, image captions or descriptors.