Published Paper Details:

This paper focuses on the digital image authentication and forgery localization using demosaicing artifacts. The aim is to make an algorithm allowing a bridge between the color filter array pattern and demosaicing algorithm estimation, and thus the statistical analysis of demosaicing artifacts in spatial domain to reinforce the authentication and localization performance. After analysing the evolution of demosaicing traces privately acquisition pipeline, a robust feature statistic characterizing demosaiced digital images is first developed on the thought of the noise residue of green channel. Such a feature statistic may be a smaller amount sensitive to the edges problem because only the graceful region of green channel is used within the event. Next, one normal mixture model is proposed to elucidate the probability distribution of feature statistics for both original and tampered images. Therefore, normality tests are often used to authenticate automatically digital images. The authentication performance are often further improved by human interpretation of supported graphic tools. Finally, a penalized expectation-maximization algorithm is used to localize forged areas in tampered images. Numerous comparative studies on four well-known datasets show that the developed algorithm yields better performance and robustness than existing forensics algorithms of the same kind