Method for retrieving massive images in real time on basis of SIFT (scale invariant feature transform) features

Abstract

The invention discloses a method for retrieving massive images in real time on the basis of SIFT (scale invariant feature transform) features. The method includes extracting the SIFT features of the images; generating SIFT genes; matching the features with one another. The SIFT genes are generated in three steps, and the three steps sequentially include generating feature information KEY; generating important feature information VALUE; compressing feature points. The method has the advantages that a feature dimension reduction process is implemented in a procedure for mapping the SIFT features to the SIFT genes, so that 128 dimensions of feature vectors are reduced to obtain 26 dimensions of the feature vectors; the feature points are compressed and processed in a procedure for forming the SIFT genes, and accordingly feature storage spaces can be greatly reduced; information of importance degrees of the features is reasonably utilized to carry out simple comparison and addition operation when the features are matched with one another, accordingly, the feature matching computational complexity is reduced and reaches O(n) level, and requirements on retrieving the massive images in real time can be met.

Description

technical field [0001] The invention relates to the field of multimedia information processing, in particular to a method for real-time retrieval of massive images based on SIFT features. Background technique [0002] Image retrieval technology has developed from text-based retrieval to content-based retrieval technology, and has made great progress, especially the introduction of SIFT features, which can be said to be the most important achievement in feature research in the field of computer vision since the 21st century. However, content-based image retrieval technology is limited by storage space and computational complexity in the face of huge, real-time expansion, and moment-changing image databases. Real-time retrieval of massive images presents great difficulties and challenges. [0003] At present, in terms of massive image retrieval, there are mainly two ways to improve retrieval efficiency. The first is to improve the search method and index method, and to improv...

AI Technical Summary

Problems solved by technology

[0006] In order to solve the problems of large image feature storage space, high feature calculation complexity, and slow retrieval speed in the current massive image retrieval, the present invention provides a real-time retrieval method for massive images based on SIFT features. The method includes the following steps:

Images