华南理工大学学报(自然科学版)

• 计算机科学与技术 • 上一篇    下一篇

基于TCS-LBP算子的视频帧复制篡改检测算法

李猛1 胡永健1† 刘琲贝1 刘海永1 廖广军2   

  1.  1. 华南理工大学 电子与信息学院,广东 广州 510640; 2. 广东警官学院,广东 广州 510230
  • 收稿日期:2017-11-15 出版日期:2018-05-25 发布日期:2018-04-03
  • 通信作者: 胡永健( 1962-) ,男,教授,博士生导师,主要从事多媒体信息安全、图像处理以及模式识别研究 E-mail:eeyjhu@scut.edu.cn
  • 作者简介:李猛( 1990-) ,男,博士生,主要从事数字视频信息隐藏的研究
  • 基金资助:
    广东省科技计划国际协同创新项目( 2017A05051002) ; 广东省自然科学基金博士科研启动项目 ( 2017A030310320) ;广州市科技计划项目( 201510010275) ;中新国际联合研究院项目( 206-A017023) 

Digital Video Frame Duplication Detection Based on TCS-LBP Operator
 

LI Meng1 HU Yongjian1 LIU Beibei1 LIU Haiyong1 LIAO Guangjun2    

  1. 1. School of Electronic and Information Engineering,South China University of Technology,Guangzhou 510640, Guangdong,China; 2. Guangdong Police College,Guangzhou 510230,Guangdong,China
  • Received:2017-11-15 Online:2018-05-25 Published:2018-04-03
  • Contact: 胡永健( 1962-) ,男,教授,博士生导师,主要从事多媒体信息安全、图像处理以及模式识别研究 E-mail:eeyjhu@scut.edu.cn
  • About author:李猛( 1990-) ,男,博士生,主要从事数字视频信息隐藏的研究
  • Supported by:
     Supported by the Science and Technology Planning Project of Guangdong Province( 2017A05051002) and the Doctoral Start-Up Fund of Natural Science Foundation of Guangdong Province( 2017A030310320) 

摘要:

针对监控视频帧复制篡改,提出一种基于时空域特征的篡改检测算法. 受空域局 部二值模式( LBP) 算子设计的启发,设计一种时域 TCS-LBP 算子,反映当前视频帧与其 前后若干帧在同一空间位置像素之间的关系; 然后在当前图像上逐点计算 TCS-LBP 特征 值,构造能同时反映当前视频帧时域和空域特征的特征图像; 利用特征图像,逐帧检测是 否存在帧复制;对于初步检测到的篡改区域,再进行虚警和漏检的修正以及篡改边界的精 确定位. 实验结果表明,文中算法具有良好的性能,与现有的2 种同类算法相比,性能明显 提升

关键词: 视频篡改检测, 时域特征, 空域特征, TCS-LBP 算子, 帧复制篡改

Abstract: A spatial and temporal feature-based detection algorithm is proposed for surveillance video duplication detection. Motivated by the design of LBP ( Local Binary Patterns) operator,we design a temporal domain CenterSymmetric Local Binary Pattern ( TCS-LBP) operator which can reflect the relationship between the pixel on the current frame and those pixels on its front and back frames in the same spatial location. By calculating the TCS-LBP value pixel-by-pixel on the current frame,we construct its correspondent feature image which can reflect both spatial and temporal features of the current frame. The feature image is thus used for detecting video duplication framebyframe. For the obtained results,we propose schemes for false alarms removal and missing detection correction. Finally,we accurately determine the boundaries of forgery region. Experimental results demonstrate that the proposed algorithm has good performance. It well outperforms the two similar algorithms in the literature. 

Key words: video forgery detection, spatial domain feature, temporal domain feature, TCS-LBP operator, video frame duplication detection

中图分类号: