Som-Based K-Nearest Neighbors Search in Large Image Databases
Zaher Al Aghbari, Kun Seok Oh, Yaokai Feng, Akifumi Makinouchi · 2002
We address the problem of K -Nearest Neighbors (KNN) search in large image databases. Our approach is to cluster the database of n points (i.e. images) using a self-organizing map algorithm. Then, we map each cluster into a point in one-dimensional distance space. From these mapped points, we construct a simple, compact and yet fast index structure, called array-index . Unlike most indexes of KNN algorithms that require storage space exponential in dimensions, the array-index requires a storage space that is linear in the number of generated clusters. Due to the simplicity and compactness of the array-index ,the experiments show that our method outperforms other well know methods. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.