边缘异常识别下视频图像篡改细节检测
Video Image Tampering Detail Detection under Edge Abnormal Recognition
陈烽 1杨怀1
作者信息
- 1. 西藏民族大学信息工程学院,陕西 咸阳 712082
- 折叠
摘要
视频图像不同栅格位置或不同压缩区域被合成为篡改图像时会出现特征块效应的差异,改变原视频的关键信息.为了准确识别图像中被篡改的像素点,提出基于边缘异常识别的视频图像篡改检测方法.通过离散余弦变换,将能量全部集中到图像的低频系数内,描述出视频的边缘等细节.利用能量比与频域熵间关系得出图像中能量的可疑度,结合预测掩膜概率图划分出发生篡改的位置区域.利用Sobel边缘检测边缘点,量化边缘点特征判断出边缘是否异常,当出现异常对其跟踪直至目标消失,检测出视频图像中的篡改位置区域.实验结果表明,所提方法能够精准检测出视频图像被篡改位置,且耗时低于 1ms,应用优势显著.
Abstract
When different grids or compressed areas in video images are synthesized into a tampered image,some differences may exist in the feature block effect,thus changing the key information.In order to accurately identify the tampered pixels in the image,this article presented a method of detecting tampered details in video images based on edge anomaly recognition.Through the discrete cosine transform,all the energy was concentrated in the low-frequency coefficients of the image in order to describe the details such as the edges of the video.Then,the suspicious level of the energy in the image was obtained by the relationship between the energy ratio and the frequency domain entropy.According to the prediction mask probability plot,the location regions where tampering occurred were divided.Moreo-ver,the Sobel edge was used to detect edge points and quantify their characteristics,and thus to judge whether their edges were abnormal.If yes,it was tracked until the target disappeared.Finally,the tampered position area in the vid-eo image could be detected.Experimental results show that the tampered position of video images can be accurately detected,and the time is less than 1ms,so this method has significant application advantages.
关键词
边缘异常/视频图像篡改/能量可疑度/预测掩膜概率图/篡改检测Key words
Edge anomaly/Video image tampering/Suspicious level of energy/Prediction mask probability plot/Tamper detection引用本文复制引用
出版年
2024
NETL
NSTL
万方数据