HOG image feature extraction algorithm based on vector homomorphic encryption

Abstract

The invention discloses a HOG image feature extraction algorithm based on vector homomorphic encryption and relates to the technical field of image processing. A VHE homomorphic encryption only supports operation between integers but does not support the division in a cryptograph domain, so based on the operation supported by the VHE, the existing HOG image feature extraction algorithm is properly cut and improved, some feature extraction steps are simplified and equivalent conversion is performed on some complex operations in the extraction processes, so under the premise that certain algorithm efficiency of the cut algorithm is ensured, the extracted feature vectors are equivalent to the feature vectors extracted by the original HOG image feature extraction algorithm, and image features can be precisely expressed. According to the invention, the adopted vector-based homomorphic encryption VHE is capable of directly encrypting an integer vector and supporting some operation calculations based on the cryptograph vectors, so compared with the former homomorphic encryption scheme based on single bit encryption or single integer encryption, the calculation efficiency in the cryptograph field is greatly improved.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a HOG image feature extraction algorithm based on vector homomorphic encryption. Background technique [0002] The Histogram of Oriented Gradient (HOG) feature is a feature descriptor used for object detection in computer vision and image processing. It forms features by calculating and counting the gradient direction histogram of the local area of ​​the image. The core idea of ​​HOG is that the detected local object shape can be described by the distribution of light intensity gradient or edge direction. By dividing the entire image into small connected regions (called cells), each cell generates a histogram of oriented gradients or the edge direction of the pixel in the cell, and the combination of these histograms can represent (the object of the detected object) description son. To improve accuracy, the local histogram can be compared and normalized by calculating...

AI Technical Summary

Problems solved by technology

[0014] The purpose of the present invention is to solve (1) the privacy report of the existing HOG image feature extraction algorithm, which is easy to be mined and analyzed by the cloud or others; (2) the gradient direction of the HOG image feature extraction algorithm is 9, and the efficiency cannot be taken into account at the same time and the problem of accuracy, the present invention provides a HOG image feature extraction algorithm based on vector homomorphic encryption, compared with the previous homomorphic encryption scheme based on single bit or single integer encryption, it greatly improves the computational efficiency in the ciphertext domain

Images