Reducing prediction error in watermarking recovery process by using majority voting

This paper proposes a method based on majority voting technique to improve the existing watermark retrieval performance [9]. Principally, the watermark retrieval process is accomplished by subtracting the predicted original pixels from the watermarked pixels to obtain the watermark bit. However, in the prediction process, prediction errors may occur and cause some errors in the retrieved watermark. Based on the assumption that errors can occur not only in the center pixel but also in any position within a prediction area, in this paper, we show that properly removing such prediction errors by changing the pixel values in the prediction area in accordance with majority vote in such a area helps improve the accuracy of the retrieved watermark. Experimental results prove that our proposed method achieves a higher accuracy of watermark retrieval. Moreover, the robustness against various types of attacks is improved as well.