Information Embedding Ensuring the Complete Recoverability of Cover Data

本研究揭示了无损数据压缩与无损数据嵌入之间的关系。无损数据嵌入是一种能够在提取嵌入数据后恢复覆盖数据的水印技术。建立了一种一维形式化无损广义lsb嵌入方法的信息论模型。在该模型的基础上，首先考虑嵌入水印所需封面数据长度的最小化问题，并设计了一种能渐近获得最小封面数据长度的通用嵌入方法。然后，我们开发了一种既不执行压缩也不执行膨胀的LZ77变体。相反，它会在保持输入文件大小不变的情况下嵌入尽可能多的额外信息。在此基础上推导了渐近嵌入容量，表明了数据压缩与无损嵌入之间的对偶关系。

This study throws light on the relationship between lossless data compression and lossless data embedding. Lossless data embedding is a watermarking technique that is capable of recovering the cover data after the embedded data has been extracted. We establish an information-theoretic model for a one-dimensional formalization of lossless generalized LSB-embedding methods. Based on the model, we first consider the problem of minimizing the Length of the cover data required to embed a watermark, and design a universal embedding method that can asymptotically attain the minimum cover data length. We then develop an LZ77 variation that performs neither compression nor expansion. Instead, it embeds as much extra information as possible while keeping the input file size. Asymptotic embedding capacity is then derived on the proposed variation, which shows a duality relation between data compression and lossless data embedding.