Modeling of remote sensing image content using attributed relational graphs
Joint IAPR International Workshops on Statistical Techniques in Pattern Recognition (SPR) and Structural and Syntactic Pattern Recognition (SSPR), 2006
475 - 483
Item Usage Stats
Automatic content modeling and retrieval in remote sensing image databases are important and challenging problems. Statistical pattern recognition and computer vision algorithms concentrate on feature-based analysis and representations in pixel or region levels whereas syntactic and structural techniques focus on modeling symbolic representations for interpreting scenes. We describe a hybrid hierarchical approach for image content modeling and retrieval. First, scenes are decomposed into regions using pixel-based classifiers and an iterative split-and-merge algorithm. Next, spatial relationships of regions are computed using boundary, distance and orientation information based on different region representations. Finally, scenes are modeled using attributed relational graphs that combine region class information and spatial arrangements. We demonstrate the effectiveness of this approach in query scenarios that cannot be expressed by traditional approaches but where the proposed models can capture both feature and spatial characteristics of scenes and can retrieve similar areas according to their high-level semantic content. © Springer-Verlag Berlin Heidelberg 2006.
Automatic content modeling
Remote sensing images
Content based retrieval