Evaluation of low-level features by decisive feature papperns

In content-based image retrieval (CBIR), the effectiveness of the low-level features depends on their capabilities in describing the high-level semantic concepts. How to properly evaluate such an effectiveness remains a challenge. In this paper, we address the evaluation problem by using the decisive feature patterns of the low-level features. Intuitively, a decisive feature pattern is a combination of low-level feature values that are unique and significant for describing a semantic concept. An evaluation study on three low-level features shows that our method can tackle the evaluation problem well. That is, the decisive feature patterns can properly characterize the low-level features' capabilities in describing the semantic concepts.