Supercharge Your Innovation With Domain-Expert AI Agents!

An image copy detection method based on generative adversarial networks

A detection method and image technology, applied in biological neural network models, instruments, character and pattern recognition, etc., can solve problems such as difficult to describe robust features, complex redundancy, complex image information expression, etc., to avoid complicated processes and limitations performance, program simplicity, and good robustness

Active Publication Date: 2019-03-29
TIANJIN UNIV
View PDF7 Cites 8 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] The above algorithm extracts image features by manual methods, these features are redundant and complex, and the information expression of the image is very complicated, especially some abstract robust features are difficult to describe

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • An image copy detection method based on generative adversarial networks
  • An image copy detection method based on generative adversarial networks
  • An image copy detection method based on generative adversarial networks

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0041] In order to improve the recognition accuracy of the image copy detection algorithm, an embodiment of the present invention proposes an image copy detection method based on a generative confrontation network, which is composed of a generator and a discriminator. The role of the generator is to generate tampered images based on the original image to fool the discriminator for copy detection, while the role of the discriminator is to distinguish the copied image pair from the non-copy image pair, see figure 1 , the specific operation is as follows:

[0042] 101: Apply distortion processing to the original image in the training set to generate a copy image, perform normalized preprocessing on the original image and the copy image, and form the original image and the corresponding copy image into a positive sample, and form different original images into a negative sample;

[0043] Among them, the normalized preprocessing of the original image and the copied image is specifi...

Embodiment 2

[0075] The method in Embodiment 1 is described in detail below in conjunction with specific parameters and calculation formulas, see the following description for details:

[0076] 201: image preprocessing;

[0077] Among them, 10 are randomly selected from the ImageNet dataset 4 image, and fix the image to the standard size of 256×256.

[0078] 202: Generate a copy image;

[0079] Apply a random distortion to the normalized image, generating 10 4 copy image. Types of random distortion include: JPEG lossy compression, Gaussian noise, rotation, median filtering, histogram equalization, gamma correction, adding speckle noise, and loop filtering. Normalize the original image and copy image pixels to zero mean, standard deviation, and then linearly transform the pixel value range to [-1,1].

[0080] 203: Construct a discriminator;

[0081] The discriminator network used in the embodiment of the present invention is an improved Alexnet network, including: five convolutional l...

Embodiment 3

[0111] The scheme in embodiment 1 and 2 is carried out feasibility verification by experimental data below, see the following description for details:

[0112] Sampling 3×10 in the ImageNet database 3 The test images do not overlap with the original images in the training set. Do the following processing on the test image:

[0113] 1) Normalize the test image to a standard size of 256×256;

[0114] 2) Apply median filtering and Gaussian noise distortions to the normalized test image, and select different degrees for each distortion, as shown in Table 1.

[0115] Table 1 Distortion types and parameter settings

[0116]

[0117] Each original image is distorted to produce 19 copy versions, forming a 5.7×10 3 copy images. Combining two pairs of original images For non-copy image pairs. Using the discriminator trained in Embodiment 2, obtain the hash distance between the copy image pair and the non-copy image pair, and perform histogram statistics. If the hash distance...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

An image copy detection method based on generative adversarial networks is disclosed. The method comprises the following steps: applying distortion processing to an original image in a training set togenerate a copy image, normalizing the original image and the copy image, forming a positive sample from the original image and a corresponding copy image, and forming a negative sample from different original images; The network parameters of the generator and the discriminator are randomly initialized, the tampered image of the original image is generated by the generator, and then the generator parameters are fixed, and the discriminator is trained by the original image, the copied image and the tampered image. Fixing the discriminator parameters and training the generator to generate a tampered image for counteracting the discriminator; Alternatively, the discriminator and generator are trained until the number of iterations is reached and the training is completed. The invention opposes the discriminator by generating an opposing sample, thereby improving the robustness of the discriminator and the detection ability to the copied image.

Description

technical field [0001] The invention relates to the field of image copy detection, in particular to an image copy detection method based on a generation confrontation network. Background technique [0002] With the development of the Internet and the popularity of smart phones, editing and storage of digital images are very convenient. The rise of content sharing networks has also broadened the channels for digital content delivery. However, the circulation of a large number of unauthorized copied images on the Internet has damaged the copyrights of content creators. Image copy detection is a key technology for copyright management of network images. [0003] Early content-based image copy detection algorithms mainly use image histograms as features. Then Bhat et al. used the spatial information of the image to divide the image into blocks and count the average gray value of the image sub-blocks, sorting according to the gray value [1]. Kim proposed to use discrete cosin...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G06K9/32G06K9/46G06K9/62G06N3/04
CPCG06V10/24G06V10/50G06N3/048G06F18/22G06F18/214Y02T10/40
Inventor 李岳楠张凯昱
Owner TIANJIN UNIV
Features
  • R&D
  • Intellectual Property
  • Life Sciences
  • Materials
  • Tech Scout
Why Patsnap Eureka
  • Unparalleled Data Quality
  • Higher Quality Content
  • 60% Fewer Hallucinations
Social media
Patsnap Eureka Blog
Learn More

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2025 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com