Implementation and Improvement of Some Digital Image Watermarking Algorithms

• Main Authors: Chun-Hsiang Huang, 黃俊翔
• Other Authors: Ja-Ling Wu
• Format: Others
• Language: en_US
• Published: 1999
• Online Access: http://ndltd.ncl.edu.tw/handle/18506317794500715519

碩士 === 國立臺灣大學 === 資訊工程學研究所 === 87 === With the rapid growth of electronic publishing industry, copyright protection of electronic documents has become quite important. Recently, digital watermark seems to be a potential solution to discourage unauthorized copying and illegal distribution of digital documents. A useful digital watermark should satisfy the following requirements:  Perceptual invisible after embedding  Robust against compression and common digital manipulation  Unambiguous after watermark extraction  Statistical undetectable In recent years, many digital watermarking techniques are developed for image data. Among these proposals, some approaches use spread-spectrum techniques, and others use perceptual or statistical models. In this paper, a Discrete Cosine Transform (DCT) based watermarking system and a Discrete Wavelet Transform (DWT) based multi-resolution watermarking system are implemented. These two systems are implemented according to [4]. In the following sections, briefings of these watermarking algorithms and performance analysis of each system will be presented. Details and difficulties during the real implementation will be included, too. In order to achieve better perceptual invisibility and robustness against various attacks, we proposed an genetic algorithm (GA) based watermark optimization algorithm for further performance improvement of the existing watermarking algorithms. In our approach, watermark embedding positions are simulated by chromosomes and several GA optimization operators such as reproduction, crossover, and mutation are used. For the purpose of easy implementation and verification, we choose the DCT based watermarking technique proposed in [4] to prove the effectiveness of our algorithm, since we have its source code. The experimental result shows that the proposed approach can effectively improve the quality of the watermarked image and the robustness of the embedded watermark against various image manipulations.