Ultra low complexity image retrieval method based on sequence preserving hashing

Abstract

The invention discloses an ultra low complexity image retrieval method based on sequence preserving hashing, and relates to image retrieval. For images in an image library, a part of images are selected randomly to serve as a training set, and corresponding image features are extracted; the dimension of an original image feature is reduced to the length same as a hash code through a nonlinear principal component analysis method; a series of supporting points are obtained through a K-means clustering algorithm to serve as the basis of follow-up hash function learning; a corresponding hash function is learned through iterative optimization; the corresponding hash function is output, and a hash code of the whole image library is calculated; for a querying image, a corresponding GIST feature is extracted, hash coding is conducted on the image feature according to the hash code function obtained through training, the hamming distance between the hash code of the querying image and an image feature code in the image library is calculated, the similarity between the querying image and a to-be-retrieved image in the image library is measured through the hamming distance, and the image high in similarity is returned.

Description

technical field [0001] The invention relates to image retrieval, in particular to an ultra-low-complexity image retrieval method based on sequence-preserving hash. Background technique [0002] With the development of Internet technology, the amount of image data in the network is increasing exponentially. How to efficiently organize, manage and analyze these data is very important. Content-Based Image Retrieval (Content Based Image Retrieval, CBIR) technology emerges at the historic moment, and has received extensive attention from academia and industry. Generally, the CBIR system can be divided into two parts: 1. Image feature expression; 2. Efficient retrieval algorithm. [0003] The feature expression of the image is to extract the relevant features of the image to describe the content of the image, such as VLAD features (for details, please refer to the article H, Jegou, M.Douze, C. Schmid and P.Perez "Aggregating local descriptors into a compact image representation ...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to solve the previous unsupervised hash learning algorithm, the training model is to limit the feature measurement space, that is, the usual model is to carry out model learning and optimization in the Euclidean space, in order to deal with and deal with large-scale image search problems, for Overcome various problems in large-scale image retrieval, improve the scope of use of the model, can handle image search problems in different feature metric spaces, and provide an ultra-low-complexity image retrieval method based on sequence-preserving hashing

Images