%O Report %F barry:AREH %A Rafkind, Barry %A Chang, Shih-Fu %T Angular Radial Edge Histogram %I Columbia University %X The Angular Radial Edge Histogram (AREH) extends prior work on a compact image representation based on geometric distributions of edge pixels. Histograms of edge pixels are computed along the dimensions of angle and radius. Fourier transform is applied to accommodate rotation invariance. It achieves invariance for scale variation, in-plane rotation, and translation. By using a grid-based approach, it can also handle limited cases of partial occlusions. Our method is efficient . it does not require image segmentation or edge linking. We have evaluated its performance using an extensive set of test data including 12 different types of variations over diverse images (schematic diagrams and photos). Experimental results in image retrieval show that AREH outperforms baseline solutions using global edge direction histograms by a large margin %U http://www.ee.columbia.edu/dvmm/publications/06/barry-WHmap-06.pdf %8 November %D 2006