欢迎访问《图学学报》 分享到：
图学学报

图学学报

• 图像处理与计算机视觉 • 上一篇    下一篇

基于空间转换网络的视频盲水印方法

  

  1. (1. 武警部队参谋部作战勤务保障大队，北京 100089； 2. 北京林业大学信息学院，北京 100083； 3. 北京电影学院数字媒体学院，北京 100088)
  • 出版日期:2020-10-31 发布日期:2020-11-05
  • 通讯作者: 邹 玲(1986–)，女，湖南长沙人，讲师，博士，硕士生导师。主要研究方向为数字图像处理与模式识别。E-mail：zouling@bfa.edu.cn
  • 作者简介:胡 海(1982?)，男，重庆人，工程师，硕士。主要研究方向为信息分析与处理、声像技术、教育技术、软件工程。 E-mail：278331037@qq.com
  • 基金资助:
    北京市教育委员会科研计划项目(KM201910050001)；国家自然科学基金项目(61902201)

Video blind watermarking via spatial transformer networks

  1. (1. Operations and Services Group of PAP Staff, Beijing 100089, China; 2. Beijing Forestry University, School of Information Science and Technology, Beijing 100083, China; 3. Digital Media School, Beijing Film Academy, Beijing 100088, China)
  • Online:2020-10-31 Published:2020-11-05
  • Contact: ZOU Ling (1986–), female, lecture, Ph.D. Her main research interests cover digital image processing and pattern recognition. E-mail：zouling@bfa.edu.cn
  • About author:HU Hai (1982–), male, engineer, master. His main research interests cover information analysis and processing, audio-visual technology, educational technology, software engineering. E-mail：278331037@qq.com
  • Supported by:
    Scientific Research Project of Beijing Municipal Commission of Education (KM201910050001); National Natural Science Foundation of China (61902201)

PDF (PC)

64

摘要： 提出并实现了一种基于空间转换网络(STN)的视频盲水印方法。针对目前视频水印 不能同时抵抗噪声、压缩等信号处理攻击和缩放、剪切等几何攻击的问题，设计一个子块水印 嵌入机制，并依赖 STN 确定水印嵌入子块区域以及空间变换系数。选择变换域中的中频系数作 为嵌入通道，通过空间变换系数矫正旋转和缩放后的图像已获得鲁棒视频水印方法。实验结果 表明，基于 STN 的视频盲水印方法具有较高的视觉隐蔽性以及较强的稳定性和鲁棒性，同时嵌 入水印容量达到 256 位以上。

关键词: 版权保护, 空间转换网络, 数字水印, 离散余弦变换, 视频盲水印

Abstract: This paper presented a novel adaptive video watermarking algorithm based on spatial transformer networks. An adaptive video watermarking was embedded into a block detected by the attention model from spatial transformer networks to resist noise and compress attacks, as well as the geometric attack. According to the attention mechanism derived from the masking behavior of the human perceptual model, a visual saliency block was extracted for further watermarking embedding. Experimental results show that the watermarked video is highly imperceptible and resistant to many degradations and distortions of video, and reaches over 256 bits in watermark capacity.

Key words: copyright protection, space transformation network, digital watermarking, discrete cosine transform, video blind watermarking