基于模式分类的图像区域作伪检测

为了鉴别一幅数字图像是否存在作伪的区域,提出一种利用改进的图像特征进行区域作伪检测的算法.基于模式分类的思想,该方法把图像分割成适当大小的块,从图像块中提取特征数据,用SVM分类器训练数据并得到支持向量机模型,利用该模型检测嫌疑图片是否存在作伪.该算法从噪声相关性、残差噪声、图像质量、小波域等方面分析相机图片的特点,获取每种的统计特征,形成特征集.实验结果表明,该方法能有效地检测出图像的具体作伪区域.

Detecting spurious area of image based on pattern classification

An novel approach for detecting spurious area of digital image is proposed by using improved image features.Based on the idea of pattern classification,the whole images is segmented into blocks of appropriate size,then feature data of image blocks is input to the SVM classifier to obtain a support vector machine model.The spurious area of the suspicious image is determined by using the model.The camera photos are analyzed from several angles,including noise correlation,noise residual,image quality and wavelet analysis,etc.And the statistical characteristics of each are obtained to form a feature set.Experimental results demonstrate that the proposed method can effectively identify the specific area of spurious images with high accuracy.