Nonlinear Approximate Indexing for Multimedia Data

This paper presents a new nonlinear approximate indexing method for highdimensional data such as multimedia data. The new indexing method is designed for approximate similarity searches and all the work is performed in the transformed Gaussian space. In this indexing method, we first map the input space into a feature space via the Gaussian mapping, and then compute the top eigenvectors in the Gaussian space to capture the cluster structure based on the eigenvectors. We describe each cluster with a minimal hypersphere containing all objects in the cluster, derive the similarity measure for each cluster individually and construct a bitmap index for each cluster. Finally we transform the nearest neighbor query into the hyperrectangular range query and search the clusters near the query point. The experimental results for our new indexing method show considerable effectiveness and efficiency.