Water chaotic fruit fly optimization-based deep convolutional neural network for image watermarking using wavelet transform

Abstract

Due to the rapid growth of multimedia in network technology, accessing the digital media becomes very easy. Hence, protecting the intellectual property requires more interest in image watermarking. For this sake, different image watermarking approaches are developed, but it poses robustness and transparency issues. Therefore, an effective image watermarking method named Water Chaotic fruit fly Optimization algorithm-based Deep Convolutional neural network (WCFOA-based Deep CNN) is developed for embedding the secret message to the cover media. The proposed WCFOA is developed by integrating the Water Wave Optimization (WWO) with the Chaotic Fruit Fly Optimization algorithm (CFOA). The inspiration of propagation operator and the refraction operator increases the diversity of population and minimizes the premature convergence. However, the breaking, propagation and the refraction operator of the proposed optimization shows the effectiveness of balance between the exploitation of exploration phase in search space using the fitness measure. Accordingly, the embedding process is achieved using the wavelet transform with the selected optimal region using the evaluated fitness value. Several images of brain tumors from BRATS dataset, with tumors having different contrast and form, are used to assess the proposed method. The experimental analysis shows that, the proposed WCFOA-based Deep CNN obtained better performance using the metrics, like correlation coefficient and Peak signal-to-noise ratio (PSNR) with the values of 1 and 45.2157 using without noise scenario and the correlation coefficient and PSNR of 0.9918 and 45.0627 for Impulse noise. By considering the salt and pepper noise, the correlation coefficient and PSNR is 0.9918 and 47.001 and in the Gaussian noise scenario the values of correlation coefficient and PSNR is 0.990 and 46.985, respectively.