本文提出一种基于多分辨率分解的可视数字水印技术。利用图像的多分辨率分解技术，相同分辨率层次的灰度级数字水印嵌入到对应的相同分辨率层次的原始静态图像之中，使水印对原始图像具有自适应性。由于水印的嵌入过程是基于原始图像的不同分辨率层次之间的关系，所以水印的提取不需要原始图像。实验结果表明本文提出的数字水印技术，经一些图像处理操作和有损压缩后仍是健壮的。

Password protected cryptosystem is commonly adopted nowadays. The main shortcoming of such cryptosystem is the lack of the connection between the user and password. The special biometric protected cryptosystem adopting biometric key scheme, which binds the biometric template and cryptographic key monolithically, can be a good solution to the problem. The biometric key scheme is devised in such a way that the cryptographic key cannot be retrieved without a successful biometric authentication. A novel biometric key scheme based on modular secret sharing is proposed in this paper. The feasibility and high security have been proved by theoretical analysis and experiments. In the experiment of combining iris recognition with the scheme, the probability for illegitimate user to retrieve the cryptographic key is zero; the probability for legal user fail to retrieve the cryptographic key is 0.0571. The optimal biometric key is 2645 times more difficult for an attacker to retrieve the cryptographic key comparing to the legitimate user.

几何攻击是目前困扰数字水印领域发展的问题，现有的抗几何攻击的水印技术存在诸多缺陷和不足。本文提出一种基于时域的抗几何攻击水印技术，利用边框这一简单的图形通过对其检测和提取，完成对几何失真参数的检测。考虑到实际应用中存在噪声的影响，用中值滤波和锐化增加了几何参数估计的鲁棒性。利用检测获得的几何失真参数对图像进行校正，使图像基本恢复原图从而无失真地提取水印。同时本文探讨了诸多文献尚未提及的几何失真恢复问题，提出一种模板匹配的方法恢复图像。仿真实验证明，该方法在含水印图像受到几何攻击后，仍可无失真地提取水印。

A method of embedding a gray-level digital watermark into an image is proposed in this paper. By stack filter's threshold decomposition technique, a gray-level digital watermark is decomposed into a series of binary digital images for implementing multiple watermarking. Experimental results show that the proposed techniques are successfully tolerant to reasonable image processing operations and the lossy compression techniques such as Joint Photographic Experts Group (JPEG) compression techniques.

A video watermarking technique based on discrete fractional Fourier transform (DFRFT) is proposed. Each frame of original video is first decomposed into two-dimensional (2-D) multiresolution representations by 2-D discrete wavelet transforms (DWT) along the spatial axis. Then the wavelet coefficient frames in each group of pictures (GOP, each is 16 frames) are transformed into DFRFT coefficient frames by one-dimensional (1-D) DFRFT along the temporal axis. The watermark is embedded into each DFRFT coefficient frame in the GOP, and the angular parameter of the DFRFT can be changed to adapt itself to the original video. Experimental results show that the proposed technique is robust enough against the attacks of frame dropping, averaging and lossy compression.