Skip to Main Content
In this paper, we propose a known-host-state methodology for designing image watermarks that are particularly robust to compression. The proposed approach outperforms traditional spread spectrum watermarking across all JPEG quality factors. The fundamental approach uses 2D chirps as spreading functions, followed by a chirp transform, to recover the watermark. Because this method can spectrally shape the chirp to match image content and JPEG quantization, its performance is greatly enhanced. The energy localization of the chirp is exploited to embed low power watermark per image blocks while maintaining reliable detection performance.