基于SURF的航行数据记录仪图像真实性鉴别

Authenticity Detection of VDR Image Based on SURF

针对使用复制-粘贴篡改手段,掩盖VDR图像中海上事故真相的问题,提出基于SURF特征提取算法的VDR图像盲取证方法。采用快速Hessian矩阵提取SURF特征点,把匹配的特征点用线连接起来,从而检测出篡改区域。与已有算法进行对比表明,该算法具有较好的检测效果,且对高斯模糊、乘法噪声和重压缩等操作具有较强的鲁棒性。

For problems about covering up the truth of marine accidents with copy-move forgery method in VDR images, a VDR image copy-move blind forensics method based on SURF feature extracting was proposed. The VDR image database was built, and forgery images were obtained by conducting copy-move forgery operations using Adobe Photoshop CS, combining with some geometric transformation. The fast-Hessian matrix was used to extract the SURF feature points, and the matching feature points were connected by lines to detect the tampered region. The method was compared with the previous methods, showing that the proposed method is successful in detecting forgery images. It is also robust to operations such as Gaussian blur, multiplicative noise, and recompression.