Image Device Source Identification Based On Pattern Recognition

Digital image can be easily generated and forgery, in this case, the reliability problems of digital image source suffered a serious question. In order to solve this problem, the digital image forensics that an objective, fair and to clarify the truth verification technology is proposed. It has an important significance to ensure the reliability of the digital image source.This paper is based on pattern classification approach to research the source of digital image from digital camera and a color scanner. Discuss the performance and robustness problem of source identification algorithm in the ideal mode and actual application mode. The ideal mode that test images are not processed, and the actual application pattern means that the test images are tampered. The contributions of this thesis are described as follows.1. An improved source camera identification algorithm based on image features is proposed. More ideal classification effect can be got by using the method. Then, we analyze the robustness of the current source identification and point out the problem in the design processing of the existing camera source identification algorithms. Finally, we give the direction to solve this problem.2. A robust source scanner identification algorithm is proposed. The proposed algorithm aims at solving the robust problem of the existing source scanner identification. It improves the accuracy of classification, but it also reduces the amount of calculation.3. Using the sequential forward floating search algorithm for choosing features. Using image feature subsets, we analyze the performance and robustness of source camera/scanner identification, and then, discussed the impact of using these subsets on the robustness of these algorithms. We also give the direction to solve the robustness problem of feature selection.