A novel refinement method for automatic image annotation systems

Nowadays, increasing demand on image annotation requirement increases the number of automatic image annotation systems. However, performances of current annotation methods are far from practical usage. Common problem of current methods is the gap between semantic words and low level visual descriptors. Because of semantic gap, annotation results of these methods contain irrelevant noisy words. To give more relevant results, refinement methods should be applied. In this work, we represent a novel refinement approach for image annotation problem. Proposed system produces candidate annotations by combining relation between words and raw annotation. Optimal one of the produced annotations is used for annotation.