Watermarking Based on Multiple Description VQ

Abstract

Considering transferring watermarked data over channels, we may encounter one question: can the signals of the embedded watermark(s) survived under the noise of the channels? In this chapter, unlike the watermarking systems introduced previously which apply artificial procedures (or attacks in other term) to test the robustness, we focus on the channel noise and introduce a hybrid watermarking scheme to overcome the losses caused by channel errors. The system is based on VQ and multiple description coding (MDC), and also has the ability of embedding dual watermarks. The VQ-based watermarking system first modifies the VQ indices to hide the first watermark, as illustrated in Chap.6, and applies the well-known traditional LSB scheme to hide the second watermark, as introduced in Chap.4. The concept of MDC is then employed to generate the considered number of descriptions, which are transmitted to the receivers via different channels. On the receiver side, accoring to the number of descriptions received, the embedded dual watermarks can be extracted by applying the corresponding inverse embedding procedures. At the end of this chapter, we present the simulation results to highlight the superiority of this scheme.