Recently

SVMs(support vector machines) have been widely used in image retrieval as a method to improve the retrieval performance. However

conventional SVMs encounter four problems: small size of positive samples

asymmetry problem of training samples

over-fitting and weakly real-time. To solve these problems

an asymmetric bagging based fuzzy support vector machine (AB-FSVM) is proposed. An asymmetric bagging is made to negative samples

and then based on fuzzy theory and SVM

the retrieval images are gotten. Experimental results based on a set of Corel images show that the proposed system performs much better than the previous methods

especially when the size of positive samples is small.