A Quantization-Based Image Watermarking Scheme Using Vector Dot Product

In this paper, we proposed a quantization-based watermarking scheme for image data. To increase the robustness of the watermark, each bit of a binary watermark string is embedded into a vector composed of several wavelet coefficients selected from low and middle frequency domain of an image. The proposed scheme is based on quantization technique, so the space of the vector is divided into several subspaces. Then, the vector is modified such that it will fall into the subspace corresponding to the value of the watermark bit. Since subspaces divided by quantization operation are distributed over the whole space, the vector could be modified so that the square error is minimized. The experimental results illustrate that the proposed method can generate a more robust watermark while keeping image fidelity.