Visible watermarking in document images using two-stage fuzzy inference system

The key problem of visible watermarking is how to balance the watermark visibility, the security and the quality of marked image. In this paper, an adaptive visible watermarking scheme in document images using two-stage Mamdani fuzzy inference system (FIS) is presented. Firstly, four attribute parameters of document images including the neighborhood gray-scale value (G), skewness (Sk), entropy value (En) and standard deviation (Std) are defined as visible watermark embedding criteria. Secondly, the FIS1 and FIS2 are designed with different input and output parameters to get the adaptive intensity factors. In order to avoid the visible watermark being removed by the binary removal attack, the gray-scale uniform distribution method is used to remove the peak of the probability logarithmic histogram after the FIS1 stage. Finally, according to the results of FIS2 stage, the change of histogram is not obvious. To evaluate and analyze the performance of this scheme, the proposed scheme is compared with other previous visible watermarking schemes, and experiment results show that the presented one has better visual effect and less distortion.