Research About Image Tampering Detection Based On Processing Traces

In today's society, digital image is an important medium which bearing the image information. Various types group. Such as the mass groups, the news media, research units and institutions, are using digital image as the important information carrier. In this case, the integrity, authenticity and other features of the digital image will be damaged and influenced. And on some occasions, such as the court, insurance compensation and scientific research, need real complete reliable image, all this make the image authentication and tamper detection become an important research direction.There are two important ways on the image authentication. One method is active detection authentication, the other method is blind detection authentication. Method of blind detection authentication apply to a wide scope, and better adapted to the current certification status of the image. This paper mainly discusses the blind detection based on blur traces of tampering image.The paper first introduces the image tampering techniques and methods, and then briefly introduces the image active and passive authentication; Then, based on the the mathematical model of the Gaussian blur, the paper presents and analyzes a variety of tamper detection methods of blur tampering. On this basis, we determine that this paper's work is the detection of tampering about the Gaussian blur operation. The paper proposed the image tamper detection algorithm aim to blur operation, using the frequency distribution of image in the frequency domain. The main work of the paper in the following aspects.1.Discussed the rudimentary knowledge of image Gaussian blur and tampering detection algorithm aim at Gaussian blur. First we introduce the basic knowledge and basic features of the Gaussian blur, then analyse and discuss of the methods proposed by the previous:wavelet decomposition method, the image color analysis methods, and the method based on the image edge properties and so on. We proposed new structure of the blur operation detection algorithm theoretical basis by summing up the advantages and disadvantages of various methods.2.Proposed a detection method, using high-pass filtering that can enhanced blur traces in the Fourier transform domain. The method uses the properties of high-pass filtering and the properties of the frequency distribution of image in the Fourier transform domain. We choose homomorphism high-pass filter to process the image in the frequency domain,to achieve the purpose of enhancing blurred region. Then we segment the image using Iteration threshold. After that we can take the image edge using mathematical morphology. This moment, we can locate the tampered image regions. The experimental results show that the method can effectively achieve image blur detection and locate the tamper region, and has a certain robustness.3.In the Fourier transform domain, we can only achieve better results on the blur edges of the image, we proposed a detection algorithm in the DCT domain that can get better result aim at block blur. On the one hand the DCT transform is closely related to Fourier transform, in fact DCT transform is the real part of Fourier transform, and have a similar frequency-domain properties with FFT transform; On the other hand DCT transform based on image block; According the image frequency properties in the DCT domain, the paper propose a blur estimation algorithm. Simulation results show that the algorithm for blur image area can get a nice detection results.