Hierarchical Image Representation Based on Digital Topology for Bridging Remote Sensing and GIS

It's necessary to discuss the topology model of digital images for integrating remote sensing and geographic information system in higher levels. Based on cellular complex theories, a hierarchical image representation is presented that maintains both raster and vector representations of an object inside a same data structure and to translate between spatial concepts in image understanding and mining. And a set of functional architectures that efficiently reflects this representation preserves topological relation equivalency between discrete and continuous spatial data, without physically merging it. In order to put these discrete-related ideas into computational practice, a novel progressive discrete segmentation algorithm is implemented that incorporates spatial information related to adjacency between pixels. The particularity of this algorithm is that connected regions and their topology generate objects in different scales, furthermore constructing a tree-object structure reflecting their spatial relationships.