poor robustness and low detection precise were induced by using local pixels to decision authenticity, and ignoring the strong spatial dependence of the source in current image forgery detection algorithms so that can not detect small real area of forgery; in addition, the high complexity and severe time consuming were existed by the idea of detection step by step adopted in theses algorithms. So the image forgery detection algorithm based on convex optimization mechanism coupled sensor pattern noise was proposed. The best priori probability model of image label pixel was constructed by introducing Markov random field and taking into account the strong spatial dependence of the source; embedding the Bayes rule, and jointing entire image pixels to balance the observed statistics and the prior knowledge on the image and penalizing maximum probability label pixel maps for detection the authenticity of the image; taking image forgery detection into convex optimization problem, and designed the convex optimization mechanism to improve the detection efficiency. Experimental results show that this algorithm had stronger robust and higher detection accuracy and efficiency.