A handful of watermarking schemes, which employ Genetic algorithm for improved performance, have been presented in the literature for protecting the copyrights of digital images. Recently, incorporating Genetic algorithm into watermarking schemes to improve its performance and effectiveness has received a great deal of attention among watermarking researchers. A brief review of some recent researches is presented here.
Minerva M Yeung 1997 proposed a technique to verify whether the image content is changed or tampered with by using an invisible watermark stamped in the image pixel values. A modified error diffusion procedure is used where pixel values are modified to produce small and random changes. A verification key is produced with the stamped image. For the extraction process the verification key is generated by pseudo random number generator which is in form of three sets of binary look up tables for a color image. The results are encouraging and show no visual artifacts introduced due to watermarking.
Fabien A P Petitcolas 1999 presented a detailed survey on information hiding. They have looked at the aspect of robustness with respect to common signal processing algorithms. Normally being considered are image compression, additive Gaussian Noise, low pass filtering, scaling and cropping. However rotations are normally ignored. Three attacks are considered-namely robustness attacks- to remove/diminish watermark, presentation attacks-modify content so that watermark is no longer detected,(Mosaic attack) and lastly interpretation attack- attacker creates a situation where ownership rights cannot be asserted. The motivation to use Stirmark to handle geometric distortions is also considered in the survey.
Mohanty S P 1999 proposed an image watermarking in the Discrete Cosine Transform domain. The embedding is based on modifying the DCT coefficients of host image by developing a mathematical model which is based on texture sensitivity of human visual system. The scaling factors and the embedding factors are chosen such that edge blocks should be least altered to avoid significant distortion of image. Although the perceptual quality of image is preserved this does not give a robust watermarking which indicates a requirement for better transform techniques combined with optimization.
G.C.Langelaar 2000 presented an article to give an overview of common watermarking techniques. For printed images dithering patterns is suggested. To embed watermark the histogram modification is suggested over the common approach of pixel value changes. Another important concept is self embedding of the cover image itself is used to detect against tampering of original image and if possible recover the content. The two classes of watermarking techniques namely correlation based and non correlation based methods are extensively considered.
Shinfeng Lin 2000 presented a novel technique for embedding watermark in frequency domain. Embedding the watermark at low frequency and using the weighted correction gives an improvement in imperceptibility of the watermark. The robustness against attacks like image compression and geometrical distortions indicate an impressive PSNR and NC values.
Jonathan K Su 2001 has given a theoretical presentation of three iterative numerical methods to optimize the watermark vector, namely Greedy Marginal Analysis(GMA) , Simulated Annealing(SA) and Greedy Annealing(GA). The results confirm that there is not a unique optimum defense for all attack distortions. The rule of thumb as per the results is that white watermarks perform well at low distortions. For high distortions PSC(power spectrum condition) –compliant watermarks perform well. The choice of watermark power spectrum is highly application dependent.
Chin-Shiuh Shieh 2003 presented the application of genetic algorithm using transform domain techniques namely discrete cosine transform. The results are impressive under various types of attacks like compression, cropping etc. With consideration to computation time, parameters in GA need to be carefully chosen to get a optimized output. It is also crucial to determine the conflicting requirements associated with their weighting factors in the fitness function before designing the GA based watermarking system.
Frank Y.Shih 2004 considered the problem of rounding of pixel values when the watermark image is of large size. This scheme is for enhancement of watermark retrieval. A conversion rule is framed for choosing the best chromosome by an adopted fitness function to convert the real numbers into integers during the transformation. Out of the popular know evolutionary algorithms like simulated annealing, hill climbing and GA’s, GA’s were found to be most suitable to solve the rounding problem.
Asifullah Khan 2004 solved the tradeoff between robustness and imperceptibility by using the evolutionary techniques to get a near optimal solution. The scheme uses the Discrete Cosine Transform combined with image adaptiveness in giving a perceptual shaping function(PSF)which will shape the watermark according to cover image. There is an improved resistance to attacks, especially against JPEG compression. However with DCT there is the drawback of suffering against geometrical attacks where performance is lowered.
Dongeun Lee 2006 presented a watermarking algorithm in the discrete wavelet transform domain using evolutionary algorithm. The proposed algorithm consists of wavelet-domain watermark insertion and genetic algorithm-based watermark extraction. More specifically watermark was inserted to the low-frequency region of wavelet transform domain, and watermark extraction was efficiently performed by using the evolutionary algorithm. The proposed watermarking algorithm was robust against various attacks such as JPEG image compression and geometric transformations.
Zhicheng Wei 2006 has implemented a GA optimization using DCT technique for embedding the watermark.GA is used to choose the AC coefficients which are modified to to embed the spread spectrum watermark. The GA adaptive tool gives a performance improvement. The selection of bands for embedding the watermark is important and varies from one block to another. As compared to Cox method, this scheme gives robust watermarking against different types of attacks.
Chin-Shiuh Shieh 2006 implemented digital watermarking using Vector Quantization domain. Genetic Algorithm is used to reassign the indices of code words. Due to this, embedded information is diffused more evenly across the image to be protected. Any security leakage is obviously avoided. For consideration of storage space and bandwidth requirement vector quantization, an image compression technique is used. VQ offers high compression rates while preserving the image content. The code book size in VQ is a tradeoff between two parameters viz compression quality and compression rate. An improved bit correct rate is clearly seen under various types of attacks in this case.
Hsiang-Cheh Huang 2007 has proposed a watermarking scheme based on progressive transmission with genetic algorithms (GAs). They implemented the watermarking embedding and extraction systems in the transform domain, and applied the JPEG spectral selection mode for scalable transmission of the watermarked image. By employing a GA with a proper fitness function into the watermarking system, both the watermark imperceptibility and watermark robustness requirements are considered and optimized. The number of embedded bits, or the watermark capacity, is much larger than that in other existing algorithms in the literature. Also, the watermark capacity in the proposed algorithm lies within the theoretical limit. In addition, the embedded watermark can be partly extracted at the receiver side even when the watermarked image is being transmitted. Simulation results showed both the robustness and the effectiveness of progressive transmission under different attacking schemes and different bandwidth variations.
Ning Zhong 2007 resolved the problem to solve two conflicting objectives: robustness and invisibility by using genetic algorithm (GA) to obtain the optimized solution. It is difficult to satisfy both of them at same time. Traditional watermarking algorithm often solve this problem by choosing parameters via experience, which is always inefficient. With a proper fitness function in the watermarking system, GA search for the optimal parameters to improve the performance of watermarking algorithm. The simulation results showed that both watermark robustness and invisibility can be achieved under different attack schemes.
Shu-Chuan Chu 2008 proposed an optimized scheme for watermarking based on zerotrees. Conventional techniques in the literature mainly perform watermark embedding and extraction processes in the transform domain, including the discrete Fourier transform, discrete cosine transform, and discrete wavelet transform domains. The three watermarking requirements above were in conflict with each other; therefore, they were finding a way to obtain a trade-off among them. They first performed watermarking in the wavelet domain. Next, they properly select zerotrees in a wavelet transform with the genetic algorithm. The simulation results not only demonstrated better performances of the watermarked images after optimization, but also reveal the robustness of the extracted watermarks under common attacks.
Veysel Aslantas 2008 presented an optimal watermarking scheme based on singular-value decomposition (SVD) using genetic algorithm (GA). The singular values (SVs) of the host image are modified by multiple scaling factors to embed the watermark image. Modifications are optimized using GA to obtain the highest possible robustness without losing the transparency.
Experimental results showed both the significant improvement in transparency and the robustness under attacks. Based on existing experiences to evaluate the applicability of robust watermarking, it is generally agreed that three parameters or requirements, including the quality of watermarked contents, the survivability of extracted watermark after deliberate or unintentional attacks, and the number of bits embedded, need to be considered. However, performances relating to these three parameters conflict with each other, and the trade off must be searched for.
Hsiang-Cheh Huang 2009 has taken all the three requirements into consideration, and add the flexibility to meet the specific design in implementation. With the aid of genetic algorithm, they designed an applicable system that would obtain the good quality, acceptable survivability, and reasonable capacity after watermarking. Simulation results presented the effectiveness in practical implementation and possible application of the proposed algorithm.
Zorana Bankovic 2009 demonstrated the effectiveness of using GA’s in fast searching of the space of the possible solutions . A high detection rate was achieved after a relatively short period of training time. Also the by retraining, the system becomes highly adaptable. As the GA’s are inherently parallel in operation there is a possibility of using reconfigurable hardware with the implementation cost much lower. At the same time GA’s can search the solution space in multiple directions at once.
Sanjeev Kumar 2009 presented a digital watermarking algorithm in discrete wavelet transform (DWT) domain for stereo image coding. First, a disparity-image was computed from the pair of stereo images using a frequency domain based matching criteria. Later, this disparity-image was used as a watermark and embedded into the degraded host (left stereo) image based on a modifying singular values concept. The host image was degraded using Arnold transform. Finally, real coded genetic algorithm (RCGA) was used to estimate the optimal order of Arnold transform and the strength of watermark to fulfill the tasks of security, invisibility and robustness in proposed algorithm. In proposed algorithm, a legal user can retrieve the embedded watermark (disparity-image) and so able to recover 3-D information and right image of the stereo- pair. Experimental results were presented to evaluate the performance of proposed algorithm in terms of accuracy and robustness.
Chih-Chin Lai 2009 presented a robust digital image watermarking scheme based on singular value decomposition (SVD) and micro-genetic algorithm (micro-GA). In an SVD-based watermarking scheme, the singular values of the cover image are modified by considering multiple scaling factors to embed the watermark image. Determining the proper values of scaling factors is not an easy task. They viewed it as an optimization problem and apply the micro-GA to efficiently obtain the values. Experimental results showed that the proposed approach has good performance against several attacks.
Chen Yongqiang 2009 presented a DWT domain image watermarking scheme to meet the watermarking properties: security, imperceptibility and robustness. In the scheme, watermark comes from a meaningful binary image encrypted by two-dimensional chaotic stream encryption that has more security. In the procedure of watermark embedding, the watermark is embedded into host image through selecting and modifying the wavelet coefficients using Genetic algorithm with a simple fitness function to improve the imperceptibility of watermarked image. In order to identify the owner of extracted watermark, Synergetic Neural Network is used in the watermarking identification to overcome the limitation of correlation analysis or the human sense organ after some attacks. The results of their scheme realization and robust experiments showed that the scheme has preferable performance.
Jiann-Shu Lee 2009 proposed a watermarking algorithm for uncompressed video based on Quantization Index Modulation (QIM) and differential energy. The Differential Energy Watermarking (DEW) algorithm has been demonstrated as an effective video watermarking algorithm;while in some scenarios, DEW algorithm cannot provide enough robustness and fidelity. This problem has been solved by above authors. The experimental results indicated that the proposed algorithm is more robust than original DEW and modified low-frequency DEW for lossy compression and transcoding, while maintaining high fidelity.
Giulia Boato 2009 presented a flexible benchmarking tool based on genetic algorithms (GA) and designed to assess the robustness of any digital image watermarking system. The main idea was to evaluate robustness in terms of perceptual quality, measured by weighted peak signal-to-noise ratio. Through a stochastic approach, they optimized this quality metric, by finding the minimal degradation that needs to be introduced in a marked image in order to remove the embedded watermark. Given a set of attacks, chosen according to the considered application scenario, GA support the optimization of the parameters to be assigned to each processing operation, in order to obtain an unmarked image with perceptual quality as high as possible. Extensive experimental results demonstrated the effectiveness of the proposed evaluation tool.
Prayoth Kumsawat 2009 proposed a digital image watermarking algorithm in the multiwavelet transform domain. The embedding technique is based on the quantization index modulation technique and this technique does not require the original image in the watermark extraction. They have developed an optimization technique using the genetic algorithms to search for optimal quantization steps to improve the quality of watermarked image and robustness of the watermark. In addition, they analyzed the performance of the proposed algorithm in terms of peak signal to noise ratio and normalized correlation. The experimental results showed that the proposed method can improve the quality of the watermarked image and give more robustness of the watermark as compared to previous works.