A New Scheme for Optimum Decoding of Additive Watermarks Embedded in Spatial Domain

This paper addresses the problem on digital watermarking, called the data hiding problem that refers to embedment and extraction of hidden information in a given image with the highest possible reliability. The performance of a watermarking system is measured by the probability of error for the extracted information. We present a new solution which significantly improves the performance of additive watermarking in the spatial domain. The embedment considered in this paper selects the watermarked regions randomly depending on a secret key while it fixes the watermark. We construct an optimum decoding scheme which uses the differences of luminance components in extraction. Then, we prove that the proposed decoding scheme significantly improves the performance. The theoretical results are confirmed experimentally.