Image Feature Layered Indexing Method for Transparent Access

Abstract

In the data transparent access process, it is often necessary to query the data accurately to find the location of the data that the user needs to access. This paper studies the image feature organization and indexing of image data based on image content retrieval, and uses the image plane position constraint relationship to propose an image feature layered index method. In the process of image transparent access, data storage management needs to detect and extract local feature information of the image (such as SIFT feature points), and then design appropriate data structures to organize and index massive high-dimensional feature vectors, and in a specific storage structure. Design an efficient and accurate feature search algorithm. The paper mainly includes two parts: LSH high-dimensional indexing algorithm for thumbnails and KD-Tree high-dimensional indexing algorithm based on improved position information.