Publications

Abstract : The increasing effect of illegal exploitation and imitation of digital images in the field of image processing has led to the urgent development in the growth of copyright protection methods. Digital watermarking has proved best in protecting illegal authentication of data. In this paper, we propose a digital image watermarking scheme based on computational intelligence paradigms like Genetic Algorithm (GA) and Particle Swarm Optimization (PSO). The input digital host images undergo a set of pre-watermarking stages like image segmentation, feature extraction, orientation assignment, and image normalization to obtain image invariance properties when subject to attacks. Expectation Maximization (EM) algorithm is used to segment the images and the features are extracted using Difference of Gaussian (DoG) technique. The feature maps from the feature extraction methods locate the magnitude by orientation assignment making the circular regions invariant. The resultant image is normalized by scaling to acquire the scaling invariance for the circular region. The watermark image is then embedded into the host image using Discrete Wavelet Transform (DWT). During the extraction process, GA, and PSO are applied to improve the robustness, and fidelity of the watermarked image by evaluating the fitness function. The perceptual transparency and the robustness of the watermarked and the extracted images are evaluated by applying filtering attacks, additive noise, rotation, scaling and JPEG compression attacks to the watermarked image. From the simulation results the performance of the Particle Swarm Optimization technique is proved best based on the computed robustness and transparency measures along with the evaluated parameters like elapsed time, computation time and fitness value. The performance of proposed scheme was evaluated with a set of 50 textures images taken from online resources of Tampere University of Technology, Finland and the entire algorithm for different stages was simulated using MATLAB R2008b.