Detection of Triple JPEG Compressed Color Images
Hao Wang, Jinwei Wang, Jiangtao Zhai, Xiangyang Luo · IEEE Access · 2019
In order to extend the detection of JPEG compressed color images to solve the real-life problem, three-class classification forensics of JPEG compressed color images with the same quantization matrix is proposed. Since the previous methods treat detection of JPEG compressed color images as binary classification and JPEG compression with the same quantization matrix leaves slight tracks, three-class classification forensics of JPEG compressed color images with the same quantization matrix is a new and challenging problem. In this paper, two aspects are considered to solve this problem. First, if images are compressed, rounding and truncation error will occur. Thus, preprocessing of images is performed to extract error to highlight statistical difference which can help to classify. Second, the support vector machine (SVM) algorithm is originally designed for the binary classification problem, so dealing with a three-class problem, it is necessary to reconstruct a suitable three-class classifier. Besides, convolutional neural network (CNN) parallelly deal with three channels of the color image. The relationship of the three channels is terminated. However, quaternion convolutional neural network (QCNN) which utilizes quaternion algebra not only is directly used to three-class classification but also retain the relationship between three channels. Experimental results demonstrate that the proposed method achieves good performance and is better than the state-of-the-art approaches investigated.