中文摘要：

数字图像被动盲取证技术是对图像的完整性和真实性进行鉴别。图像遭受篡改操作后，为了消除图像伪造在拼接边缘产生畸变，伪造者通常会采用后处理消除伪造痕迹，其中，模糊操作是最常用的手法之一。因此提出了一种针对人工模糊的取证方法。首先，利用非抽样Contourlet变换分析图像边缘点特征进行边缘点分类；然后通过统计正常边缘点与模糊边缘点之间的差异鉴别模糊边缘；最后引入局部清晰度来区分人工模糊与离焦模糊，从而最终标定人工篡改边缘痕迹。实验表明该方法能够有效地检测出图像人工模糊篡改操作，较为准确地定位图像篡改边界。伪造图像边缘模糊越严重方法的检测效果越好。与其他模糊检测方法相比所提方法具有像素级别定位能力。

英文摘要：

Digital image passive blind forensics aim to distinguish the integrity and authenticity of digital images. After the image is tampered, in order to eliminate the visual edge distortion caused by splicing during the process of forgery, some post-processing operations are usually utilized to eliminate the tampering traces. For example, manual blurring is one of the common approaches. Based on this condition, a method which can detect manual blurring from the tampered image is proposed. Firstly, the features of the image edges are analyzed by using nonsubsampled contourlet, by which the image edges can be classified. Then the authors can distinguish whether the edge is blurred by the differences between the normal edge and the blurred edge. Finally, local definition which is defined to indicate the differences between the manual blurring and out of focus, can be used to locate the manual blurred traces. Experimental results show that this method can detect possible manual blurring in the given tested images and locate the tampered boundary with a relative high accurate rate. The more serious the image blurred edge is, the better performance the method has. Compared with the other blurring detection approaches, the method presented has the capability of pixel-location.