Informatics Publishing Limited

Sawsan M. Gharghory ¹, Hanan H. Elazhary ¹

Affiliations
1 Computers and Systems Department, Electronics Research Institute, Cairo, EG

Abstract

Digital image watermarking is a technique in which a secret watermark is embedded into the host image to make an assertion about the host image ownership. A practical image watermarking algorithm should be transparent, robust, secure, blind, and fast. Unfortunately, there is hardly any algorithm in the literature with all these characteristics. In this paper we propose a novel algorithm that satisfies all these properties of digital image watermarking. The proposed technique is a grayscale digital image watermarking using novel visual cryptography and cluster computing. In contrast to other algorithms in the literature, the proposed technique constructs two new shares, the size of each share is equal to that of the watermark and their generation is straightforward and fast. The algorithm embeds a given watermark into the Discrete Cosine Transform (DCT) domain. It is robust since it utilizes the relative values of the low frequency components rather than the absolute values. It is transparent since the watermark bits are not physically embedded into the host image. The algorithm is secure due to the use Torus Automorphism (TA) permutation and a pseudo random number generator. It is also blind since few keys and two shares are needed for watermark extraction. Finally, the algorithm relies on cluster computing for significant speedup. The experimental results prove the robustness of the proposed technique against several types of unintentional attacks. Besides, the results prove the practicality of our approach comparing with other techniques in literature.

Keywords

Digital Image Watermarking, Visual Cryptography, Security, Cluster Computing, High Performance Computing.

Full Text

     

Hanan H. Elazhary ¹, Sawsan M. Gharghory ¹

Affiliations
1 Computers and Systems Department, Electronics Research Institute, Cairo, EG

Abstract

Particle Swarm Optimization (PSO) algorithm is used in the literature to solve different types of optimization problems. Most PSO algorithms suffer from the problem of the long processing time and the fact that PSO gets trapped easily in a local minimum. This paper argues the need to address both problems simultaneously to enhance the performance of PSO unlike other research studies in the literature that typically address only one of these problems. To emphasize this, the paper proposes to combine the basic PSO algorithm with both evolutionary operators and parallel processing to solve the rounding error problem of a DCT-based fragile image watermarking algorithm. Evolutionary operators help PSO jump out of local minima, while parallel processing helps speed up processing. In such watermarking algorithms, the watermark bits are typically embedded by modifying the least significant bits of the frequency coefficients of the DCT-transformed host image. The host image is then converted into the spatial domain using inverse DCT and the real numbers of the inverse-DCT coefficients are rounded to integers. This enables detecting the slightest modification of the host image, but the rounding process results in a significant difference between the extracted watermark and the embedded one and reduces the watermarked host image quality. The proposed algorithm is compared to three cases: a) using PSO alone, b) using PSO with evolutionary operators without parallel processing, and c) using PSO with parallel processing and without evolutionary operators. The experimental results show the superiority of our approach.

Keywords

Cauchy Mutation, Evolutionary Operators, Fragile Image Watermarking, Parallel Processing, Particle Swarm Optimization.

Full Text