The Research Of Image Authentication Methods Based On Quantization Watermark

With the rapid development of Internet and Multimedia technology, it is very easy to copy, edit and modify digital images. Illegal tamper and transmission will cause critical harm to society and people, so multimedia content protection and authentication is becoming urgent in the world. Digital watermark can overcome the limitations of traditional authentication methods and thus become very hot in image authentication research.This thesis targets at the image authentication method based on quantization watermark. It first gives a thorough introduction of digital watermark and image authentication, including the research status quo and development trend of robust watermark, fragile watermark as well as semi-fragile watermark. It also describes the classification and methods of image authentication, and analyzes its requirements and performance evaluation.On the basis of general quantization watermark model, the relationship between robustness, embedding distortion, quantization parameter, noise intensity and distribution of host signal is analyzed. The analysis shows that the error probability mainly depends on the ratio between standard deviation of noise and quantization step size. So we can employ the above analysis results and choose optimal quantization parameters according to the specified watermark embedding distortion and sensitivity to noises in real applications. Experiments show the choosing parameter improves the algorithm performances.A semi-fragile watermark based on halftone is proposed for image content authentication. Not only can it distinguish malicious attacks from unintentional attacks, but also it can recover the tampered area. Experiments show its effectiveness.Adaptive quantization watermark is introduced to enhance the robustness to multiplication noises according to the characteristics of Human Visual System. Since the quantization step changes synchronously with the multiplication noises, this algorithm is robust to multiplication noises in principle. Experiment results prove that adaptive quantization watermark is superior to its uniform quantization counterpart.