Method mapping image low-level features to high-level semantics

Existing image management systems normally retrieve images based on low-level features. Users usually have a more abstract notion of what will satisfy them, so there is a gap between content-based system organization and the concept-based user. Based on Support Vector Machine learning, a method mapping low-level features to high-level semantics was proposed, and experiments were carried out on an especial image database, which realized image retrieval and semantic annotation.