Correction of Insertions and Deletions in Selective Watermarking

Common forward error correction is designed to correct substitution errors only. There is no detection of insertions/deletions. Such systems are usually employed in digital watermarking schemes. If watermark data is embedded selectively based on content features where de-synchronization can occur during extraction, common error correction can result in bursts of errors. But there are a few techniques that are able to handle this kind of error. We analyze these techniques, which can be separated into three approaches, namely concatenated coding, dynamic programming, and punctured channel coding. As demonstrated, the latter one fails to correct de-synchronization in second generation watermarking schemes, if the number of selected embedding locations is much smaller than the number of host signal samples. In this paper, we propose a new method that outperforms all other methods presented so far concerning insertion/deletion error correction in second generation watermarking schemes.