Heuristic relevance learning for web image annotation

Automatic annotation can automatically annotate images with semantic labels to significantly facilitate image retrieval and organization. Traditional web image annotation methods often estimate specific label relevance to image, and neglect the relevance of the assigned label set as a whole. In this paper, A novel image annotation method by heuristic relevance learning is proposed. Label-to-image relevance and label-to-label correlation are formulated into a joint framework. Measures that can estimate the relevance are designed, and the assigned label set can provide a more precise description of the image. To reduce the complexity, a heuristic algorithm is introduced, thus making the framework more applicable to the large scale web image dataset. Experimental results demonstrate the general applicability of the algorithm.