Semantics in Content-based Multimedia Retrieval

This contribution investigates the content-based feature extraction methods used in visual information retrieval, focusing on concepts that are employed for the semantic representation of media content. The background part describes the building blocks of feature extraction functions. Since numerous methods have been proposed we concentrate on the metaconcepts. The building blocks lead to a discussion of starting points for semantic enrichment of low-level features. The second part reviews features from the perspective of data quality. A case study on content-based MPEG-7 features illustrates the relativity of terms like “low-level,” “highlevel” and “semantics”. For example, often more semantics mean just more redundancy. The final part sketches the application of features in retrieval scenarios. The results of a case study suggest that – from the retrieval perspective, too – “semantic enrichment of low-level features” is a partially questionable concept. The performance of classification-based retrieval, it seems, does hardly depend on the context of features.