3.3 Relevance feedback algorithm
RF is a powerful interactive technique used for
improving the performance of the retrieval algorithm.
In some systems for content-based image retrieval[20],
the semantic concept of the query is captured and
gradually improves retrieval precision. RF can be
easily applied to MR[12, 13], thus achieving greater
improvement. With user-provided relevant/irrelevant
information on the ranking list, the goal of this
algorithm is to spread relevance by MR, and then
to more correctly and gradually improve retrieval
precision.