Language universal font watermarking with multiple cross-media robustness.

• A font-based text watermarking scheme is proposed to achieve robustness to cross-media distortions such as print-camera shooting, screen-camera shooting, and print-scanning. • A glyph centroid modification-based font codebook generation algorithm is proposed to automatically create target-similar fonts. • The proposed method does not require the participation of OCR to identify glyphs, which can improve the extraction accuracy and efficiency. • The proposed method is applicable with various font sizes, languages, and devices. The rapid development of digital devices and the increasing demand for copyright protection and leakage tracing of documents result in new demands for robust text watermarking. It becomes more important for a utility text watermarking scheme to have robustness in cross-media channels since the camera shooting operation to papers or screens has become very convenient. To achieve the corresponding cross-media robustness, this paper proposes a font-based text watermarking scheme, which is applicable to most commonly used languages while the previous works were usually designed for specific languages. To generate fonts with high efficiency, we propose a novel glyph centroid modification-based font generation algorithm, which can automatically create target-similar fonts. And we design an effective watermarking scheme that utilizes the relative centroid position (RCP) of glyphs to represent watermark signals. Compared with existing font-based text watermarking schemes that artificially design fonts, the proposed scheme ensures higher efficiency by generating the font codebook automatically. In addition, the proposed RCP-based watermarking scheme can achieve the robustness against lossy compression and several kinds of cross-media distortions. The watermark message can be extracted without the need to recognize the semantic information of glyphs by optical character recognition (OCR). [ABSTRACT FROM AUTHOR]