The Research of Energy-Efficient Watermarking Algorithms

• Main Authors: Yu-Ting Pai, 白御廷
• Other Authors: Shanq-Jang Ruan
• Format: Others
• Language: en_US
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/80451617224853737700

博士 === 國立臺灣科技大學 === 電子工程系 === 97 === Digital watermarking is a potential method for copyright protection. In order to achieve robustness, transparency, blindness, and etc., much literature has discussed how to develop a high quality watermarking. Although these previous methods offer the satisfactory results for resisting many attacks, the complicated frameworks cost a lot of area and energy consumption for hardware implementation. Since the chip area and battery life of embedded systems are always restricted, these high complexity watermarking methods are not popular used nowadays. In this dissertation, we are concerned with energy-efficient and robust watermarking technologies. We analyze the computational complexity for three type watermarking algorithms. First, we propose an improvement of basic watermarking algorithm. The frequency coefficients are exploited directly as the criterion for the permutation mapping. The energy and time consumption can be reduced without calculating variance parameter in spatial domain. Second, in order to further reduce power and achieve the requirement of blind extraction, we continue our study based on DS-CDMA based wateramrking algorithm. According to the properties of DCT and DS-CDMA, the inverse discrete cosine transform is qualified to substitute for both forward and inverse transform in proposed method. Thus, we can save quantities of energy and time consumption. Third, the study is focus on overcoming geometric attacks for video. The proposed algorithm provide the robustness via hide watermark in DCT domain, and supply the capability of resisting geometric attacks by embedded a group of template to DFT domain. For power saving, the watermark massage is embedded in DCT domain which is decoded from video directly. In additional, we propose partial DFT and downsampling methods to reduce the huge computational complexity. To validate the results, we employ Wattch toolset developed at Princeton University to evaluate power dissipation and execution time. Moreover, the quality measuring results show that the proposed watermarking algorithms is robust enough against several signal distortions.