使用自适应阈值的图像篡改检测与定位算法

为了实现对机载移动目标的快速捕获和粗跟踪瞄准,设计了粗跟踪演示系统,完成了外 场飞行实验的 初步验证。本文系统利用GPS数据完成对目标的捕获,通过对姿态数据的校正,方位误差降 到0.60°(1σ),俯 仰误差降到0.40°(1σ),有效缩小了不确定区域；系统还对跟踪算 法进行了优化改进,利用分段式函数等效 非线性调整函数,有效解决动态目标跟踪时快速调整和超调之间的矛盾。飞行实验表明, 本文的粗跟踪演示 系统的捕获时间优于10s,粗跟踪精度优于480μrad,为精跟踪子系统实现最终的目标精确跟踪瞄准提供了 有利条件,实验结果验证了该系统用于激光通信链路快速建立的可行性。

Image tampering detection and localization algorithm using adaptive thresholding

As the existing image tampering detection algorithms based on camera s ensor pattern noise have high false alarm rate in strong texture areas,a novel detection algorithm using adap tive thresholding adaptive to the texture complexity is proposed.According to Nyman Pierson criteria,the correla tion decision thresholds for different texture complexity are determined. Furthe rmore,the relationship between the correlation threshold and the texture complexity is obtained. While detecting,the correlati on coefficients of each pair of non-overlapping blocks from the noise residual of the test image and the refere nce sensor pattern noise of the inspected camera are computed first.The tampered areas are roughly localized ba sed on the correlation matching determined by the adaptive thresholds.Then,the fast zero mean normalized cross c orrelation (ZNCC) algorithm is used to calculate the correlation of the corresponding pixels in the roughly localized regions between the reference sens or pattern noise and the noise residual of the test image,and to realize the accurate localization.The experi mental results on the mobile phone image database show that the detection rate of the proposed algorithm is 98.8%,while the false alarm rate is only 1.897%.Compared with the existing methods using a f ixed threshold,the proposed algorithm using adaptive thresholds can effectively reduce the false alarm rate in the complex texture re gion ,and can accurately locate the tampered region.At the same time,the proposed algorithm can improve the effici ency by 26times compared with the conventional sliding-window-based algorithms.