Scholarlink

<p>With the continuous development of deep neural networks, more and more network models are applied to the field of image region tampering detection. However, the application of deep neural networks in video region tampering detection has just started. This paper proposes a method for tampering detection of video regions based on deep neural network. This method combines the characteristics in the spatial and time domains to extract weak features of tampered videos. The algorithm uses the multi-scale receptive field to extract the weak features of the spatial tampering of the video frame, and combines the RPN to judge and locate the tampering target. At the same time, the algorithm combines with MRNet [<CitationRef CitationID="CR1">1</CitationRef>] to extract the weak features of the video adjacent frame difference image sequence in the time domain. We tested the publicly tampered video dataset. The experimental results show that our algorithm can better identify and locate the tampered regions of the video compared to the currently known video region tampering detection algorithm.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A novel video region tampering detection method based on deep neural network

  • 期刊论文

  • 发表日期：

    2026年1月27日

查看全文
  • Hao Chen,
  • Qi Han,
  • Qiong Li,
  • Xiaojun Tong,
  • Xin Bao

摘要

With the continuous development of deep neural networks, more and more network models are applied to the field of image region tampering detection. However, the application of deep neural networks in video region tampering detection has just started. This paper proposes a method for tampering detection of video regions based on deep neural network. This method combines the characteristics in the spatial and time domains to extract weak features of tampered videos. The algorithm uses the multi-scale receptive field to extract the weak features of the spatial tampering of the video frame, and combines the RPN to judge and locate the tampering target. At the same time, the algorithm combines with MRNet [1] to extract the weak features of the video adjacent frame difference image sequence in the time domain. We tested the publicly tampered video dataset. The experimental results show that our algorithm can better identify and locate the tampered regions of the video compared to the currently known video region tampering detection algorithm.