Double Compression Analysis And Tampering Localization Of JPEG Images

With the development of information technology, digital image has become an important form of media for it is fast, easy to store and transport. At the same time, varieties of digital image processing software are more and more powerful and simple to use. Thus, tampering an image is not sophisticated. Lots of fake images have emerged in the field of political, military, judicial, etc. These forged images have influenced the opinion of viewers believing the credibility of the image they are watching. In recent years, digital image forensics has been improved.Compression history analysis of digital image is an important part of digital image forensics, which is usually used to make a raw judgment of the authenticity of images. Tampered image is often resaved in JPEG format. If the original image is in JPEG format, the tampered image will suffer two JPEG compressions (double JPEG compression) after resaving. Most existing literature focus on how to distinguish single JPEG compression image form double JEPG compression image. The few of them concern the shift of two JPEG compressions.In this thesis, we studied the JEPG compression framework and the blocking artifacts caused by double JEPG compression. Then we proposed a method to estimate the shift between two JPEG compressions. Based on this proposed method, we designed a model to locate tampered region in JPEG image. The main contribution of this thesis includes:(1) This paper introduced existing digital image forensics tools by category, including methods detecting particular image tampering processing, analysis of imaging processing and compression history. The challenges of the existing literature were discussed.(2) This paper analyzed the influence brought by the shift of double JPEG compression and established the relationship between information loss and the shift of double JPEG compression. A new method was proposed to estimate the shift of double JPEG compression. In the experiments, the proposed method outperformed the latest relative method, especially on small size image.(3) A JEPG image tampering region locating model was designed based on the proposed algorithm to estimate the shift of double JPEG compression. The tampered JPEG image can be divided into two parts, the tampered region and the background region. The two regions both suffered double JPEG compression, but the shifts of the two regions were different. Thus, we can locate the tampered region by finding uncommon shifts of double JPEG compression. This model was proved effective to locate the tampered region in the experiments. Also, it was robust to resizing, rotating, feathering to the tampered region. Compared to other methods, the proposed tampering locating method provided better result when dealing with small image size. It was more robust and didn’t need image dataset to train a machine learning based classifier or to get a proper threshold.