A Bayesian framework for automatic concept discovery in image collections

This paper develops a Bayesian framework for automatic hidden semantic concept discovery to address effective semantics-intensive content-based image retrieval. Each image in the database is segmented to regions associated with homogenous color, texture, and shape features. By employing self-organization map learning, a uniform and sparse region-based representation is obtained. With this representation a probabilistic model based on the statistical-hidden-class assumptions of the image database is developed, to which expectation-maximization technique is applied to analyze semantic concepts hidden in the database. An elaborated retrieval algorithm is designed to support the probabilistic model. The semantic similarity is measured through integrating the posterior probabilities of the transformed query image, as well as a constructed negative vector, to the discovered semantic concepts. The proposed approach has a solid statistical foundation and the experimental evaluations on a database of 10,000 general-purposed images demonstrate its promise of the retrieval effectiveness.