基于统计噪声水平分析的图像拼接检测

图像拼接是最常用的图像篡改操作之一,针对篡改图像噪声水平不一致性的现象,本文提出了一种基于统计噪声水平分析的图像拼接检测方法。首先,将检测图像分割成大小相同的非重叠图像块,然后,利用一种非参数估计算法来估计每个图像块的噪声值,并且采取聚类法对图像块的噪声值进行聚类,聚类结果分为可疑部分和非可疑部分两大类。最后,通过一个由粗到细的两阶段策略对篡改区域进行定位。哥伦比亚未压缩图像拼接检测评估图像库的实验结果表明,本文方法能够准确地估计图像块的噪声和定位出拼接区域,性能优于现有方法。

Image splicing detection based on statistical noise level analysis

Image splicing is one of the most commonly used image tampering operations.In this paper,we propose an effective image splicing detection method based on image noise level inconsistency.First,the suspicious image is divided into non-overlapping blocks.Then,the noise value of each image block is estimated by a new nonparametric algorithm.The method of clustering was used to classify according to the noise value and the clustering result was divided into two categories:suspicious part and non-suspicious part.Finally,the two-stage strategy from coarse to fine further locate the tampering area.Experimental results over Colombian Uncompressed Image Splicing Detection Evaluation Dataset demonstrate that the proposed method can detect and locate the tampered regions more accurately than state-of-the-arts investigated.