摘要
为了鉴别一幅数字图像是否存在作伪的区域,应用数字图像与数码相机的噪音相关性,提出一种检测数字图像区域作伪的新方法.该方法首先利用小波滤波器从数字图像中提取出残留噪音,采用平均多幅图像残留噪音的办法得到相机的噪音参考模式.然后选择一个适当大小的检测模块在嫌疑图像的残留噪音和相机的噪音参考模式上同步移动,同时求出两者的噪音相关系数.最后把这一噪音相关系数与预设的判别阈值进行比较,从而判定该嫌疑图像是否存在作伪区域.在假设错误接受率为10-3的条件下,利用Neyman-Pearson判别规则求出各相机的判别阈值,利用浮动阈值法和加入三个修正措施后进行检测.结果表明,该方法的区域检测正确率为87%左右.
A new method was proposed to detect spurious area of digital image; The residual noise from an image was extracted by wavelet filter,and a reference pattern of noise for a digital camera was obtained by averaging the residual noise from multiple images. A proper detector was chosen to move synchronously at both the residual noise from suspicious image and the reference pattern of noise of camera, and the correlation between the residual noise and the reference pattern of noise was computed synchronously. The spurious area of the suspicious image could be determined by comparing the correlation with a predetermined threshold. The Neyman-Pearson approach was used to derive the threshold on the assumption that the false acceptance rate is 10 3, and detects suspicious image after using the method of floating threshold and subjoining three approaches of shaping, the result shows that the correct detect rate is 87 % approximately using this method.
出处
《光子学报》
EI
CAS
CSCD
北大核心
2008年第10期2108-2113,共6页
Acta Photonica Sinica
基金
公安部重点攻关计划(2005ZDGGQHDX005)
福建省自然科学基金(A0640014)资助
