Discovering local cooccurring patterns from aerial images

Publication Type:
Conference Proceeding
Fourth International Conference on Information Technology and Applications, ICITA 2007, 2007
Issue Date:
Full metadata record
Developing a spatial searching engine to enhance the search capabilities of large spatial repositories for GIS update has attracted more and more attention. Existing methods are usually designed to extract limited types of objects and use only one aspect of features of Images. In this paper, we propose to use the local cooccurring patterns to disclose the cooccurring relationships among each dominant local features and use this local cooccurring patterns to recognize an object from aerial images. For this purpose, we investigate three types of local features: colour-based features, texture-based features, and edgebased shape features. In order to facilitate the feature extraction procedure, we first use discontinuity-preserving smoothing methods to filter the input image. Two popular smoothing techniques are tested and compared. Experimental results are presented in this paper.
Please use this identifier to cite or link to this item: