Index Structures for High-Dimensional Image Features

Most image or video search engines operate by extracting and storing feature vectors from the multimedia objects. These feature vectors may, for example, consist of colour or texture histograms and can vary considerably in size, say from a few ten numbers to 50,000 numbers. When an image database is queried with a particular example image ("show me similar images"), the corresponding feature vector is computed and the most similar feature vectors from the database are searched to display the most similar images in the database. This paper documents the design and implementation of a high dimensional index application to facilitate the speedy searching in feature based image information retrieval. We follow the structure and algorithms of the X-tree, which is designed for high dimensional indexing. Having presented these algorithms, we also purpose our improvement for the k-nearest neighbour query algorithm. Finally we present performance evaluations to show the merit and demerit of our approach.