Study Of Robust Digital Image Watermarking Based On Genetic Algorithm

With the widespread use of Internet and the development in computer industry, great convenience is provided for information storage. As a byproduct, the digital multimedia information without permission can be copied and redistributed easily from the Internet, which renders great loss to both copyright owner and producer. So copyright protection has become a serious issue. As a potential measure to solve the problem, digital watermarking is becoming a hot topic in the research area. During the transmission of digital product, some data processing is the prevalent, such as JPEG, filter, noise and so on. Since all the processes are threatening the security of the watermarked product, in order to protect the watermark effectively, the research on robust watermarking scheme is important to both the research field and the business application of digital watermarking.By combining the Genetic Algorithm (GA), applying the inner character of digital image, and taking advantage of existed research, some robust digital watermarking methods are proposed against different attacks. We hope it can provide a new way for the full application of the digital watermarking scheme.First, We propose an adaptive genetic algorithm(AGA) for digital watermarking scheme. The adaptive crossover and mutation operators have been proposed to increase the global research and decrease the complexity. The ability of global optimaztion is proved by theoreticial derivation and experiments. The Fitness function has been designed to avoid setting the weights of different parameters.Second, adaptive genetic spread sprctrum watermarking(GSSW) model in 8×8DCT domain has been made based on the classical spread sprctrum watermarking method. The GA is used to optimize the best watermark imbedding bands for improving the security, fidelity and robustness. The relative embedding method is changed to absolute method, and we design the fitness function only including robustness. Experiments show that the robustness is improved obviously and the random charater of GA make the watermarking method more secure.Third, three robustness improved subsampling watermarking schemes are introduced after we research the key parameters of subsampling blind watermarking algorithm, such as watermark insertion sequence order, transmission model, insertion bands and insertion strength. 1. After researching on the fidelity and robustness under the influence of watermark insertion order sequence, we propose a genetic subsampling watermarking method(ISW-1) where GA is used to optimize the most suitable sequence to avoid the"mesh"effect, which is produced by fixed, regular sequence or randomized sequence. 2. We also research on the characters under attacks of the subsampling watermarking method, and find the"not integrity attacks"weakness of the method. So we propose a new packet transmission subsampling blind watermarking scheme(ISW-2) in DWT domain to overcome the shortness. 3. The SPIHT based genetic subsampling blind watermarking algorithm(ISW-3) is applied in three level DWT domain. We focus on using the SPIHT method to select the perceptually most significant spectral components of the data where the watermarks is embedded by adaptive strength which is searched by GA. Keeping the same image qulity, the above three algorithms apparently preserves superior robustness under various attacks, such as JPEG, lowpass filter, Gauss noise and so on.Finally, we research on another kind of watermark——binary watermark, which can be detected by eyes. A symmetrical genetic lossless blind watermarking(SGLBW) is proposed where new polarities of the DCT coefficients and watermark insertion and extraction methods are designed to avoid the watermark leak and save about 1/3 time. The binary watermark is permuted by Arnold transform, then is embedded into the important coefficients which is optimized by GA. The method optimizes both the fundamentally conflicting requirements: robustness and invisibility.