A DCT-domain blind watermarking system using optimum detection on Laplacian model

This paper presents a digital watermarking system in the DCT domain. The watermarking system is designed to satisfy strictly the perceptual constraints. Under the constraints, robust and fragile watermarking techniques are designed by varying the watermark strength. The optimum detection structure is constructed based on the Laplacian model, which has the properties that high performance can be achieved for a large variety of statistical models, and the computation is efficient. The detection does not need the original images. It can also detect the tampered regions very accurately. The simulations validate the watermarking system. The system can be used in copyright notification and protection, broadcast monitoring and tracking, and authentication and tamper-proofing.