An Unsupervised Hash Retrieval Method Based on Clustering Feature Directions

Abstract

The invention discloses an unsupervised hash retrieval method based on clustering feature directions, belonging to the technical field of image retrieval. First extract appropriate features for different data sets; then use K-means clustering to classify similar features into one category, and take the mean of a series of similar features to become new features; use K-means for each dimension Means clustering generates two category centers, quantifies the data of each dimension according to the distance from the center, and obtains a binary code. Finally, the given query image is coded according to the above steps, and the similarity with the training set data is compared by calculating the Hamming distance to obtain the retrieval result. The invention can perform unsupervised image retrieval more efficiently and accurately, and has great practical value as a reasonable reference.

Description

technical field [0001] The invention belongs to the technical field of image retrieval, and in particular relates to an unsupervised hash retrieval method based on clustering feature directions. Background technique [0002] A typical feature of the Internet age is the dramatic increase in the amount of multimedia information. Facing large-scale image datasets, how to quickly and accurately retrieve images of interest has become an important research issue in the field of computer vision. Traditional methods for calculating image similarity are often difficult to apply to large-scale image datasets due to the high cost of calculation and storage of images. Therefore, hash algorithms for large-scale dataset retrieval emerged as the times require. [0003] At present, the unsupervised hash retrieval technology mainly includes two steps of feature dimension reduction and quantization encoding. However, the current feature dimension reduction technology applied to hash retrieval...

AI Technical Summary

Problems solved by technology

[0008] Aiming at the problem of unbalanced variance data generated by the PCA dimensionality reduction technology in the existing hashing method, and maintaining the similarity of the original space as much as possible, the present invention provides a method based on clustering the feature directions. Supervised hash retrieval method, through the transformation of the feature dimension reduction step in hash retrieval, realizes the balance of the variance of the data after dimension reduction, maintains the effect of the original spatial similarity, and has the advantages of easy operation and high retrieval accuracy Advantage

Images