基于最小二乘孪生支持向量机的图像拼接检测

图像篡改最基础的手段便是拼接,为了恢复人们对数字图像的信任,图像拼接检测变得非常重要.论文提出一种基于最小二乘孪生支持向量机的图像拼接检测算法,算法对待测图像进行对偶数复小波变换以获取不同的子带图像,对子带图像提取其马尔科夫状态转移概率矩阵,将该概率矩阵作为拼接特征向量送入最小二乘孪生支持向量机训练以获取预测模型,最后根据该模型来判断待测图像是否经过拼接.在哥伦比亚大学无压缩图像拼接检测评估库和哥伦比亚大学图像拼接检测评估库上分别进行实验,与传统算法做对比,实验结果充分证明论文所提算法具有更高的拼接检测准确率.

Image splicing detection based on least squares twin support vector machine

Image splicing was one of the most basic methods of image tampering.In order to restore people's trust in digital images,image splicing detection became more and more important.In this paper,an image splicing detection algorithm based on least squares twin support vector machine was proposed.The detected images were decomposed by using dual tree complex wavelet transform to get different sub-band images.Markov state transition probabilities matrix was extracted from those sub-band images as splicing feature.The least squares twin support vector machine was established to predict whether the image was spliced or not.The experimental results in Columbia uncompressed image splicing detection evaluation dataset and Columbia image splicing detection evaluation dataset showed that the proposed algorithm had a higher accuracy on splicing detection than that of the traditional algorithm.