Scholarlink

<p>With the proliferation of high-resolution cameras and sophisticated editing software, video manipulation has become more prevalent, raising concerns over content authenticity. This study introduces an efficient technique to detect and localize frame duplication and copy-move forgeries. The method combines perceptual hashing (pHash) for rapid similarity estimation and Scale-Invariant Feature Transform (SIFT) for precise feature matching. Initially, keyframes are extracted using KATNA, guided by the energy factor (EF) curve to target potentially altered segments. This selective processing reduces computational load by avoiding analysis of the entire frame set while preserving detection accuracy. The proposed method is fundamentally grounded in analytical linear algebraic principles and does not rely on any training phase or parameter learning process. Experimental evaluation demonstrates a detection and localization accuracy of <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(96.5\% \pm 0.4\)</EquationSource> </InlineEquation>, confirming the method’s effectiveness for video forensics and digital content verification while maintaining low computational complexity.</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

An Efficient Analytical Frame Duplication Detection using Energy Factor Analysis and Frame Hashing in Videos

  • 期刊论文

  • 发表日期：

    2026年4月10日

查看全文
  • P. Priyanka,
  • M. Baburaj

摘要

With the proliferation of high-resolution cameras and sophisticated editing software, video manipulation has become more prevalent, raising concerns over content authenticity. This study introduces an efficient technique to detect and localize frame duplication and copy-move forgeries. The method combines perceptual hashing (pHash) for rapid similarity estimation and Scale-Invariant Feature Transform (SIFT) for precise feature matching. Initially, keyframes are extracted using KATNA, guided by the energy factor (EF) curve to target potentially altered segments. This selective processing reduces computational load by avoiding analysis of the entire frame set while preserving detection accuracy. The proposed method is fundamentally grounded in analytical linear algebraic principles and does not rely on any training phase or parameter learning process. Experimental evaluation demonstrates a detection and localization accuracy of \(96.5\% \pm 0.4\) , confirming the method’s effectiveness for video forensics and digital content verification while maintaining low computational complexity.