Study Of Blind Detection Algorithm For Digital Image Tampering

With the advent of image editing software, digital image can be easily manipulated and altered. If the image is tampered maliciously, the consequence may be very serious. So the detection of image tampering has become more and more important. The traditional method to identify the authenticity of images is to embed digital watermarks or signatures in digital images in advance, but it still has a lot of weaknesses. So the blind detection for digital image tampering without digital watermarks or signatures has become a new research direction.In this paper, two algorithms of detecting digital image tampering are studied. They are based on the study of some inherent statistical characteristics of digital images.(1) A detection algorithm based on color filter array characteristic. Most digital cameras employ a single sensor and a color filter array, and then interpolate the missing color samples to obtain a color image. The interpolation introduces specific correlations and the correlations may be destroyed by the image tampering. EM algorithm can be used to analyze the correlations and detect the digital image tampering. Through the improvement of the existing model, this method speed up convergence of EM algorithm and reduce the detecting time.(2) A detection algorithm based on the statistical characteristics of double JPEG compression. At present a majority of digital images are stored in JPEG format. When creating digital forgeries, an image is loaded into the image editing software and converted to BMP format from JPEG format. After some manipulations are performed, the image is re-saved in JPEG format, so the tampered image has been double JPEG compressed. Double JPEG compression amounts to double quantization of the block DCT coefficients, which introduces specific artifacts visible in the histograms of these coefficients. So the analysis of the DCT coefficients histograms and its Fourier spectrum can be used to detect the traces of image tempering.The experimental results show that the two algorithms are both effective in the detection of digital image tempering.