图像异常边缘篡改检测及可信度评价算法

Image abnormal edge tampering detection and credibility evaluation algorithm

针对传统图像篡改检测算法适用性有限,只能检测某一类篡改区域边缘平滑处理的问题,提出一种图像异常边缘篡改检测及可信度评价算法。采用非线性预测编码去除图像像素间的相关性,降低边缘像素点与其周围像素亮度差异,从而达到弱化边缘的目的。对处理后的图像利用边缘增强算法增强真实边缘。对比图像处理前后边缘变化,寻找异常边缘点,通过异常边缘概率计算图像可信度评价图像的真伪。实验结果表明,在未知伪造操作的情况下,该算法能够较好地检测出图像中的异常边缘,并给出图像的可信度评价。

In view of the limited applicability that traditional image tampering detection algorithms can only detect a certain class of tampering region edge smoothing,proposed an algorithm for image abnormal edge tampering detection and credibility evaluation.Nonlinear prediction coding was used to remove correlations between image pixels and to reduce the brightness difference between edge pixels and their surrounding pixels,so as to weaken the region edges.The edge enhancement algorithm was adopted to enhance the real edges.The edges changes before and after the image processing were compared.Anomalous edge points were searched.Anomalous edge probability was calculated to identify the image authenticity.Experimental results indicate that the proposed algorithm can detect the abnormal edges of the image and gains its credibility under the unknown forgery operation.