Generalized lossless data hiding by multiple predictors

Abstract

Digital watermarking is to prove the ownership and the authenticity of the media. However, as some applications, such as medical and military, are sensitive to distortion, this highlights the needs of lossless watermarking. In this paper, we propose a new lossless data hiding algorithm by using multiple predictors, which extends and generalizes our previous watermarking idea (Yip, 2005). By using different predictors with different characteristics, we can choose the embedding location to be low variance region or high variance region. The PSNR and the payload capacity are high and there are no "salt-and-peppers" artifacts.