A compact deep CNN feature indexing method based on SIFT embeddings

Abstract

The invention discloses a compact deep CNN feature indexing method based on SIFT embedding, which includes: performing energy-based sparse processing on the deep CNN features of images; adopting a distance measurement optimization method, using images in CNN features and SIFT features To update the SIFT feature of the image and the CNN feature after sparse processing, so that the CNN feature after sparse processing contains the image context information based on SIFT feature, and complete the compact deep CNN feature index based on SIFT embedding. The method disclosed by the invention is used for efficiently storing the features of pictures in a huge number of databases, and effectively reducing the time required for online retrieval.

Description

technical field [0001] The invention relates to the field of image retrieval in multimedia technology, in particular to a compact deep CNN feature index method based on SIFT embedding. Background technique [0002] In the field of image retrieval, how to reliably and efficiently store and index image features in databases is a crucial issue. In large-scale image retrieval problems, in order to provide high-quality retrieval results, a huge number of database pictures is usually required, and the pictures in the database are stored in the form of features, usually each picture has more features, Or there are a small number of high-dimensional features, so the number of features that need to be saved is very considerable. In addition, efficient storage can also effectively reduce the time required for online retrieval. In the process of image retrieval, when determining the correlation ranking between images in the database and retrieved images, it is necessary to calculate ...

AI Technical Summary

Problems solved by technology

First, the sparsity of such CNN features is relatively low, making the indexing efficiency (including feature storage efficiency and calculation efficiency during query) relatively low

Second, this CNN feature is closer to the semantic level, ignoring the underlying visual characteristics, thus affecting the retrieval accuracy

Images