像素预测误差耦合似然映射的图像伪造检测算法

为了解决当前图像伪造定位技术因使用了CFA 插值，易形成颜色插值噪声而降低分辨率，导致其难以检测微小篡改区域，使其伪造检测精度较低等不足，本文提出了像素预测误差耦合似然映射的图像伪造检测算法。首先，分析颜色滤波阵列CFA插值模型，并从图像中提取绿色分量；随后，嵌入权重因子，构造预测误差及其权重方差计算模型；根据预测误差与贝叶斯理论，定义伪造特征统计模型，识别出趋于零的特征值；最后，根据特征统计模型，建立其似然率模型，输出伪造映射，完成检测。仿真结果表明：与当前图像伪造定位机制相比，本文算法拥有更强的鲁棒性，能识别定位出微小伪造像素；且拥有更高的AUC值与理想的ROC曲线。

Copy-move image forgery detection algorithm based on pixel prediction error coupled likelihood mapping

In order to solve these defects such as low forgery detection precision and difficult to detect small tampered area induced by producing color interpolation noise with low resolution for using the color filter array in current image forgery location technologies, the image forgery detection algorithm based on pixel prediction error coupled likelihood mapping was proposed in this paper. Firstly, the green component is extracted from the image based on the color filter array CFA; then the prediction error and its local weighted variance model were constructed by embedding the weight factor; and the model of the false feature statistics was defined by prediction error and Bayes theory to identify the feature value that tend to zero; Finally, the likelihood ratio of the feature was built by the feature statistical model to output forgery mapping for finishing the detection. Simulation results showed that: this algorithm had better robustness to identify small false pixels, and it had higher Area under the ROC curves value and ideal ROC curve.