数字媒体版权认证的图像水印算法与检测技术研究

RESEARCH ON THE KEY ALGORITHMS AND DETECTION TECHNOLOGY OF DIGITAL IMAGE COPYRIGHT AUTHENTICATION

本文提出了基于组合优化的数字水印嵌入方法,提高了水印算法的鲁棒性.基于QIM的数字水印嵌入算法虽然复杂度较小、鲁棒性较好,但是QIM算法未考虑对比度掩蔽特性,使用的是固定量化步长,导致此算法在局部区域的不可见性较差且对数值测度的变化非常敏感.为改进这一缺陷,本文根据Watson视觉模型采用自适应量化步长的方法,能随量化值变化而发生相应的变化.实验结果表明,该水印算法能在不可感知的前提下,对于一些常见的攻击(剪切、50%缩放、椒盐噪声和中值滤波等)具有较好的鲁棒性.

In this paper,we propose an optimization method which improved the adaptability of watermarking algorithm. Although QIM algorithm with less complexity can keep a good robustness and be used in blind detection,it lacks analysis of masking properties of contrast and the fixed quantization step makes it bad performance at invisibility of local region and sensitive to the variation of numerical measure. To improve this defect,this paper adopts adaptive quantization step based on Watson visual model,which can change with quantization value,thus make the PSNR value of watermarking image near to the ideal value of PSNRs.The results of simulations and data of this algorithm show that this watermark algorithm takes on good performance in imperceptibility and robustness in some common attacking tests（ such as cut,50% scalling,impulse noise and median filtering）.