一种基于数字水印与深度学习的图像版权保护方法  

An Image Copyright Protection Method Based on Digital Watermarking and Deep Learning

作　　者：许宪东[1] 李艳波[1] 王建伟[1] 曲天伟[1] 王亚东[1] Xu Xiandong;Li Yanbo;Wang Jianwei;Qu Tianwei;Wang Yadong(School of Computer Science and Technology,Heilongjiang Institute of Technology,Harbin,China)

机构地区：[1]黑龙江工程学院计算机科学与技术学院,黑龙江哈尔滨

出　　处：《科学技术创新》2023年第9期45-48,共4页Scientific and Technological Innovation

基　　金：黑龙江省哲学社会科学研究规划项目(19TQE312)。

摘　　要：数字水印在声音、图像、视频等多媒体数字作品的版权保护方面有着广泛的应用。其中不可见图像水印的研究主要集中在如何解决水印的鲁棒性和不可见性等问题。而对于实际应用中用户需要从大量受保护图片中提取多种数字水印,并进行分类和识别这一问题研究较少。针对这一问题,将深度学习应用到图像水印分类,通过建立水印分类神经网络实现了图像水印的分类,实验表明是一种有效的方法。Digital watermarking is widely used in the copyright protection for multimedia digital works such as sound,image,and video.Invisible image watermarking mainly focuses on how to solve the problems of robustness and invisibility.However,users have to extract multiple digital watermarks from a large number of protected images,and carry out classification and recognition in practical applications,while there less research on this issue.To solve the problem,proposed a method that can apply deep learning to image watermark classification,which can achieve image watermark classification by establishing a neural network for watermark classification,Experiments show that it is an effective method.

关 键 词：数字水印 鲁棒性 深度学习 神经网络 不可见性 

分 类 号：TP309.7[自动化与计算机技术—计算机系统结构]