Hierarchical web image classification by multi-level features

The hierarchical image classification of web content is still an open issue. In this paper, we address the problem of image classification by using not only low-level perceptual features but also high-level semantics features. We focus on the robustness and efficiency of image classification by different categorization methods on different feature sets. Our experiments reveal some characteristics in the hierarchy classification based on textual and visual features. We propose a hierarchical threshold strategy based on data structure for multi-class categorization. The evaluation results are reported and discussed.