Study of Neural Network and Channel Coding Applications in Watermarking System

• Main Authors: You-Chang Tsai, 蔡佑昌
• Other Authors: Wan-Ching Chen
• Format: Others
• Language: zh-TW
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/57686774792312217104

碩士 === 清雲科技大學 === 電機工程研究所 === 93 === Data hiding is one of the most important topics in network security and digital signal processing researches. Digital watermarking has been shown as a valuable data hiding technique. Although many efforts have been done, the ideal way to extract reliable watermark from noise-corrupted images still needs further exploring. In this thesis, varied types of noise are used to attack the image to investigate the performance of watermarking system. Experiment results show our watermark system has better noise resistance ability. A robust watermarking process that combines the techniques of channel coding and neural networks is proposed. The watermark tenacity can be enhanced by the fault tolerance ability of the coding schemes. After convolutional coding, watermarks are rearranged by the data permutation and blocks cutting processes. Next, the coded watermark data are learned by neural networks and embedded into the discrete cosine transformed frequency coefficients of the original image. Since pixels in an image have highly correlation between each other, extracted watermark may not be recognized due to heavy noise corruption. To solve this defect, a superior version of our watermarking system is presented. The noise tolerance ability is obviously raised. Experiment results demonstrate the efficacy of the novel coding watermarking methodology.