Region-based supervised annotation for semantic image retrieval

Automatic image annotation is the key to semantic-based image retrieval. We formulate image annotation as a multi-class classification problem under the multi-instance learning framework, which deals with the weak annotation problem and works with image-level ground truth training data. The relationship between low-level visual features and semantic concepts is found by supervised Bayesian learning from positive bags. For each region in the test image, a posterior probability for each concept is calculated from class densities estimated from the training set and then the probability is modified using relevance with the other regions in the image. The image-level posterior probabilities are obtained by combining the regional posterior probabilities and keywords are selected according to their ranks. The proposed algorithm is tested on standard datasets and achieves good annotation performance.