Relevance feedback in content-based image retrieval: Some recent advances

Various relevance feedback algorithms have been proposed in recent years in the area of content-based image retrieval. This paper presents some recent advances: first, the linear and kernel-based biased discriminant analysis, BiasMap, is proposed to fit the unique nature of relevance feedback as a small sample biased classification problem. As a novel variant of traditional discriminant analysis, the proposed algorithm provides a trade-off between discriminant transform and density modeling. Experimental results indicate that significant improvement in retrieval performance is achieved by the new scheme. Secondly, a word association via relevance feedback (WARF) formula is presented and tested for unification of low-level visual features and high-level semantic annotations during the process of relevance feedback.