A framework for feature-based indexing for spatial databases

We propose a data-driven method for efficient retrieval of objects as well as similar shapes with a given query object in the spatial database environment. The idea is to find some features from the image object in order to build an index search structure. For the sake of similarity matching among shapes, the features must be invariant to rotation, translation and scaling. We propose a set of generic features that are invariant to these transformations. Each feature in the feature vector is associated with a weight based on the application, which is used in the search process. Any multidimensional point access method can then be used to build an index. In this paper, a variant of the K-D-B tree is used to construct the index structure. Finally, we define a similarity measure to find objects similar to a given query object, and discuss how similarity queries can be processed using the index structure