A Reversible Digital Watermarking Algorithm for Vector Maps

A reversible digital watermarking algorithm of 2D vector maps based on BP neural network is proposed in this paper. The watermarks were converted to binary bits, and were manipulated as the coefficients of neural network; for embedding the watermarks invisibly, the weights of the neurons would be adjusted to fitting to the neural coefficients, which picking the watermark information. In the phase of watermark extraction, those coefficients were extracted from vector maps; with above trained neural network, the watermark would be output by inputting the obtained coefficients of the neurons. Experimental results show that the maps could be recovered lossless. Compared with other classical algorithms of the kind on the tests, it is indicated that the algorithm has higher robustness, better invisibility. Especially, it can resist on the common map operations: rotation, translation.