A texture-based image retrieval system for browsing
In this article, we will demonstrate that texture could large-scale aerial photographs is presented. The salient be used to select a large number of geographically salient components of this system include texture feature exfeatures including vegetation patterns, parking lots, and traction, image segmentation and grouping, learning building developments. Using texture primitives as visual similarity measure, and a texture thesaurus model for features, one can query the database to retrieve similar fast search and indexing. The texture features are computed by filtering the image with a bank of Gabor filters.
image patterns. Much of the results presented are with This is followed by a texture gradient computation to airphotos, although a similar analysis can be applied to segment each large airphoto into homogeneous regions.
LANDSAT and SPOT satellite images. A schematic dia-
A hybrid neural network algorithm is used to learn the gram of the prototype system is shown in Figure 1. This visual similarity by clustering patterns in the feature is currently being integrated into the Alexandria Digital space. With learning similarity, the retrieval performance improves significantly. Finally, a texture image thesaurus Library (ADL) project (Smith, 1996), whose goal is to is created by combining the learning similarity algorithm establish an electronic library of spatially indexed data, with a hierarchical vector quantization scheme. This theproviding Internet access to a wide collection of geosaurus facilitates the indexing process while maintaining graphic information. A significant part of this collection a good retrieval performance. Experimental results demonstrate the robustness of the overall system in searchincludes maps, satellite images, and airphotos. For examing over a large collection of airphotos and in selecting ple, the Maps and Imagery Library at the UCSB contains a diverse collection of geographic features such as over 2 million historically valuable aerial photographs. A housing developments, parking lots, highways, and airtypical airphoto can take over 25 MB of disk space, and ports.
providing access to such data raises several important issues, such as multiresolution browsing (Strobel,