Multimedia binary coding method based on supervised multi-perspective discretization

Abstract

The invention relates to a multimedia binary coding method based on supervised multi-perspective discretization. The method comprises the following steps that: S1: assuming a training set formed by npieces of images, obtaining a Hash function which contains one series of sub-functions through learning, mapping two different modals of features of a sample to an optimized feature space, obtaining one series of Hash values corresponding to Hash sub-functions, and then, converting the Hash values into a binaryzation Hash code through binary quantization; S2: obtaining a Hash function based on supervised training: defining a linear multi-classification model, optimizing the model function, and adopting a minimum squaring loss as a target function; S3: obtaining a Hash function based on a minimum quantization loss: assuming the feature of one modal, and optimizing to the minimum quantization loss through the Hash function; S4: obtaining a Hash function based on a multi-perspective anchor chart: constructing the anchor chart, and adopting an anchor chart regularization Hash function; and S5: optimizing an algorithm. By use of the method, the similarity of data in an original space can bekept, and retrieval accuracy can be improved.

Description

technical field [0001] The invention relates to the field of information storage and retrieval, in particular to a multimedia binary encoding method based on supervised multi-view discretization. Background technique [0002] A large amount of semi-structured and unstructured data is continuously generated on the Internet, making how to effectively store these data and reduce storage space consumption has become an urgent problem to be solved. The Cross-media Hashing method provides an effective way to solve the above problems. First, the cross-media hashing method encodes multimedia data into a string of fixed-length binary codes (0 / 1 or -1 / 1), which can greatly reduce the storage space of the data; Or calculate the feature distance by comparing the Hamming distance between the hash codes. [0003] The performance of cross-media retrieval mainly depends on the quality of the learned hash codes. It is usually assumed that the quality of the hash code lies in whether it ca...

AI Technical Summary

Problems solved by technology

[0005] The invention provides a multimedia binary encoding method based on supervised multi-view discretization, aiming to solve the problem of efficient hash code generation for multimedia information storage and retrieval

Images