Reversible Watermarking is a technique which can reversibly embed data to the image. There are many watermarking techniques for color and gray-scale images. However, the binary images have only one signal bit in every pixel; therefore, this makes binary image watermarking very challenging.
There are not so many techniques of reversible watermarking for bina-ry images. Until now, there are mainly three methods: Pair-Wise Logical Computation method (PWLC), run-length method, and pattern substitu-tion method (PS). Among them, the PS method has the best performance. In this dissertation, the techniques of the reversible data hiding for binary image are introduced, and then three methods based on the PS method are proposed.
The PS method has two embedding modes: the overlapping embedding and the non-overlapping embedding. In the PS method, the size of the location map (LM) sometimes is bigger than the capacity. Therefore, even no data can be embedded.
The first proposed method is called as the PFR method. It tries to reduce the size of the LM for the non-overlapping PS method. The experiment shows that the PFR method can get smaller LM than the non-overlapping PS.
The second method is called the improved PS. This method finds that the PS method sometimes fails to decode in the overlapping mode, and shows three counterexamples. It improves the PS method, so that it can be used to the overlapping mode. The experiment shows that the improved PS can decode successfully in the overlapping mode.
The selection of the pattern pair is highly related to the performance of the PS method, however, all the PS-related methods have to set the pat-tern pair before the data hiding, and it never changes throughout the data hiding process. The third proposed method adaptively selects the patten pair according to the context, so it is no need to set the pattern pair be-fore data hiding. In order to apply the PFR method to the overlapping PS method, two modes of the embedding mechanisms are proposed. More-over, three-round embedding mechanism are proposed to compress the LM The experiment shows its superiority with comparison with the previous methods.
In the section of application, it shows three applications of the binary reversible watermarking, and finally analyses the security for the application of tamper localization.