Evaluation of Performance of Watermarking Techniques and Development of its Support System

【○!1】关于对水印技术性能的评价以及利用评价方法对水印技术进行改进：本研究旨在建立一种评价水印方法性能的工具。重点研究了基于量化和基于相关的频域水印方法，并提出了如何评价这些方法的性能。在该评价方法中，我们对嵌入水印的数量、水印图像的质量、水印检测率以及对图像处理和攻击的鲁棒性进行了衡量。此外，我们可以知道水印应该嵌入在哪里，以使水印方法对图像处理和攻击具有鲁棒性。利用这些信息，我们可以提高水印方法的鲁棒性。2】基于数字信号处理技术的水印系统分析及其在鲁棒性水印系统设计中的应用：在本研究中，我们重点研究了静态图像的水印，并讨论了基于量化和基于相关的水印方法。我们首先提出了水印嵌入和提取过程的模型，并研究了水印系统对图像处理和攻击的鲁棒性。在此基础上，阐明了图像处理过程中产生的水印失真以及水印提取过程中产生检测误差的原因。然后从水印图像的质量、嵌入水印数据的数量和水印系统的鲁棒性等方面对水印系统的性能进行了讨论，并给出了一种评价系统性能的方法。此外，我们还研究了水印提取过程的改进，设计了一个反卷积滤波器，试图消除水印失真的影响。该设计采用自适应反褶积技术实现。数值实验表明，基于dct的水印系统具有良好的性能。少

【○!1】 On the evaluation of performance of watermarking techniques and the improvement of watermarking techniques using the evaluation method : In this research, we aim to establish a tool for evaluating the performance of watermarking methods. We have focused on the quantization-based and correlation-based watermarking methods in the frequency domain, and proposed how to evaluate the performance of these methods. In this evaluation method, we measure the quantity of embedded watermark, the quality of watermarked image, the rate of watermark detection, and the robustness against image processings and attacks. Furthermore, we can know where watermark should be embedded in order for a watermarking method to be robust against image processings and attacks. Using this information, we can improve the robustness of the watermark method.【○!2】 Analysis of watermarking systems using digital signal processing technique and its application to design of robust watermarking systems : In this researc … More h, we have concentrated on the watermarking of still images and deal with the quantization-based and correlation-based watermarking methods. We have first presented a model of the watermark embedding and extracting processes and investigated the robustness of the watermarking system against image processings and attacks. Based on the results, it has been clarified how distortion of the watermark occurs by image processing and the reason why the distortion causes detection errors in the watermark extracting procedure. And then we have discussed the performance of the watermarking systems from the view point of quality of the watermarked image, quantity of the embedded watermark data, and robustness of the watermarking system, and given a method for evaluating their performance. Furthermore we have studied an improvement of the watermark extracting process by designing a deconvolution filter that attempts to undo the effect of watermark distortion. The design has been achieved using the deconvolution technique with adaptive algorithms. Numerical experiments with the DCT-based watermarking system have shown good performance. Less