An Efficient Query Mining Framework Using Spatial Hidden Markov Models for Automatic Annotation of Images

A novel method for automatic annotation of images is used with keywords from a generic vocabulary of concepts or objects combined with annotation-based retrieval of images. This can be done by using spatial hidden Markov model, in which states represent concepts. The parameters of this model are estimated from a set of manually annotated training images. An image in a large test collection is then automatically annotated with the a posteriori probability of the concepts. This annotation supports content-based search of the image-collection through keywords. The keyword relevance can be constructed using Aggregate Markov Chain (AMC). A stochastic distance between images based on their annotation and the keyword relevance are captured in the AMC is then introduced. Investigation has been made in the Geometric interpretations of the proposed distance and its relation to a clustering in the keyword space. Keywords— Markovian semantic indexing, spatial hidden markov model, image annotation, query mining