Metadata Generation and Retrieval of Geographic Imagery

Novel approaches are needed to support content-based retrieval on large-scale geographic image databases. In this paper, we present a new approach termed Keyblock for content-based geographic image retrieval, which is a generalization of the text-based information retrieval technology in the image domain. In this approach, methods for extracting comprehensive geographic image features are provided, which are based on the frequency and correlation of representative blocks, termed keyblocks, of the geographic image database. Keyblocks, which are analogous to index terms in text document retrieval, can be constructed by exploiting various clustering algorithms. By comparing the performance of our approach with conventional techniques using color feature and wavelet texture feature, the experimental results demonstrate the effectiveness of our approach for geographic image retrieval.