Multi-label image annotation via Maximum Consistency

Image annotation is a challenging but important task to understand digital multimedia contents, which by nature is a multi-label classification problem because each image is usually associated with more than one semantic keyword. Exploiting the label correlations borne in multi-label classification, we propose a novel Multi-Label Maximum Consistency (MLMC) approach to seek the optimal configuration of the image similarity graph with maximized label assignment consistency. Promising results in empirical studies on three benchmark multi-label image data sets have demonstrated the effectiveness of our approach.