A Novel Graph-based Image Annotation with Two Level Bag Generators

Image annotation has been an active research topic in recent years. However, labels are usually associated with images instead of individual regions in the training set, which poses a major challenge for learning strategy. In this paper, we formulate image annotation as a semi-supervised learning problem under multi-instance learning framework. A novel graph based semi-supervised learning approach to image annotation using multiple instances is presented, which extends the conventional semi-supervised learning to multi-instance setting by introducing two level bag generator method. The experiments over Corel images have shown that this approach outperforms other methods and is effective for image annotation.