A complementary SVMs-based image annotation system

A novel automatic image annotation system is proposed, which integrates two sets of SVMs (support vector machines), namely the MIL-based (multiple instance learning) and global-feature-based SVMs, for annotation. The MIL-based features are obtained by applying MIL on the image blocks. They are further input to a set of SVMs for finding the optimum hyperplanes to categorize training images. Similarly, global color and texture features are fed into another set of SVMs for categorizing training images. Consequently, two sets of image features are constructed for each test image and are respectively sent to the two sets of SVMs, whose outputs are incorporated to obtain the final annotation results. Our system is validated using COREL images and outperforms the peer systems in terms of efficiency and accuracy.