一种新的多特征融合视频篡改检测方法  

New Detection Method of Video Tamper Based on Feature Fusion

作　　者：陈雪艳[1] 时黛[1] 

机构地区：[1]内蒙古民族大学机械工程学院,内蒙古通辽028000

出　　处：《电视技术》2015年第23期98-100,109,共4页Video Engineering

基　　金：国家自然科学基金项目(61440041)

摘　　要：三网融合的视频防插播、防篡改技术已成为信息安全领域重点攻关的关键技术。针对多媒体业务被恶意篡改的问题,提出了一种新的多特征融合的视频篡改检测方法:结合LM算法对BFGS进行融合改进,提出改进的BFGS神经网络,通过提取视频特征平均梯度和信息熵,采用改进的BFGS神经网络进行融合篡改检测。实验结果表明,该算法收敛速度快,学习精度高,能有效地对视频篡改进行检测。Technology of video anti-cutting and anti-tampering of Triple play has become the critical technology of key research in information security field. Aiming at the problem of multimedia business is malicious tampered, a new detection method of video tamper based on feature fusion is proposed： in combination with LM algorithm to improve the BFGS, a modified BFGS neural network is put forward. By extracting average energy gradient amt information entropy, tbis improved BFGS neural network is applied to tamper detection. The experimental results show that the algorithm has a fast convergence speed, high learning precision, it can detect the tampered video accurately.

关 键 词：视频篡改 特征选取 BFGS算法 

分 类 号：TN949.6[电子电信—信号与信息处理]