A novel algorithm combined with asymmetric and adaptive Bayesian feedback

Relevance feedback is an important part in content based image retrieval. The semantic gap can be reduced by relevance feedback. The result of image retrieval can be improved effectively. An integrated adaptive asymmetric feedback algorithm is proposed based on Bayesian theory. As the asymmetry of positive and negative samples, we apply different strategies to positive and negative feedback appropriately. We use the various links between the feedbacks to process the positive feedback by memory. On the other side, we design an novel method to select additional negative examples to resolve the problem of rarity of examples, which makes the fitting of conditional probability density function is more accurately. The experiments showed that the efficiency of our algorithm is better than other algorithms of feedback.