Cross-modal data discrete hash retrieval method based on similarity maintenance

Abstract

The invention discloses a cross-modal data discrete hash retrieval method based on similarity preservation. The method comprises the following steps of establishing a cross-modal retrieval data set composed of samples containing two modalities, and dividing the cross-modal retrieval data set into a training set and a test set; establishing an objective function for keeping similarity between modalities and similarity in modalities, and solving the objective function through a discrete optimization method to obtain a hash code matrix; learning a Hash function of each mode according to the Hashcode matrix; calculating Hash codes of all samples in the training set and the test set by using a Hash function; wherein one modal test set serves as a query set and the other modal training set serves as a retrieval set, calculating the Hamming distance between the Hash codes of the samples in the query set and the Hash codes of the samples in the retrieval set, wherein the sequence serves as aretrieval result. According to the method, the similarity between modalities and the similarity in the modalities can be effectively kept, the discrete characteristics of the Hash codes are considered, a discrete optimization method is adopted for solving the objective function, and therefore the cross-modality retrieval accuracy is improved.

Description

technical field [0001] The invention relates to a cross-modal retrieval method in the technical field of multimedia retrieval, in particular to a discrete hash retrieval method for cross-modal data based on similarity preservation. Background technique [0002] With the rapid development of Internet information technology, multimedia information in various modes on the network is growing explosively. Following the development trend, cross-modal retrieval has become a crucial issue and has attracted the attention of many researchers. A typical scenario of cross-modal retrieval is to retrieve a query sample of a modality and retrieve other modalities similar to it. However, due to the heterogeneity gap, it is impossible to directly measure the similarity between different modalities. In addition, due to the explosive growth of data, it is necessary to pay attention to the storage cost and efficiency of large-scale retrieval. The hash method is a very popular method in recen...

AI Technical Summary

Problems solved by technology

However, this relaxation strategy can affect the retrieval performance

Images