Graph-Based Features for Image Retrieval
Caihua Li, Zhe‐Ming Lu · 2011
This paper proposes a novel kind of graph-based features for image retrieval. For each color image, we divide it into R, G, B component images. For each component image, we view the 256 gray-levels as nodes and construct the Gray-level Co-occurrence Graph (GCG) by counting the number of occurrences for each possible gray-level pair as neighbors in the image. Based on the generated three directed weighted graphs GCG_R/G/B, we use the in-degree histograms (IDH), out-degree histograms (ODH), in-strength histograms (ISH) and out-strength histograms (OSH) for image retrieval. Experimental results show that our features outperform the traditional color histogram-based features in terms of Precision-Recall (P-R) curve.