Channel coding in video watermarking: use of soft decoding to improve the watermark retrieval

Watermarking is often modeled as the transmission of a message over a very noisy channel. Indeed, the watermark power must be very low to ensure invisibility, and the modifications suffered by an image can be rather strong (JPEG or MPEG compression for instance), leading to a high noise level. To ensure a reliable transmission on a channel with such a low SNR, channel coding is thus mandatory. However, as standard codes are not adapted to such noisy channels, we propose a coding scheme based on concatenated codes. We show that, for these codes, optimal decoding is far too complex, and that hard decoding algorithms used in practice lead to poor performance. As soft decoding enables to approach optimal decoding at lower cost, we adapt the classical Chase (1972) decoding algorithm to the watermarking context, and propose a new soft decoding algorithm based on tree exploration