Digital Watermarking Using Machine Learning

Digital watermarking is a technique used for the information of the images that provides security for the confidentiality. The repetitions of the multimedia objects (i.e. audio, video, text, etc.) have been protected by some of the developed digital watermarking techniques. Digital Watermarking is the process of concealing messages in digital contents in order to verify the rightful owner of the copyright protection. In this paper we have proposed a method that would assist its users to embed a watermark to the cover image based on an adaptive approach in a much robust way while maintaining the quality of the cover image. The implementation of this algorithm is based upon cascading the features of DWT and PCA using Bhattacharyya distance and Kurtosis. PCA decompose and compress the watermark, which results in better PSNR and NCC values for the tested images. The proposed algorithm uses Bhattacharyya distance and Kurtosis to detect the scaling and embedding factors making it adaptive to the input image rather than providing constant value. Also, we take one step forward by investigating the construction of feed-forward denoising convolutional neural networks (DnCNNs) to embrace the progress in very deep architecture, learning algorithm, and regularization method into image denoising. Specifically, residual learning and batch normalization are utilized to speed up the training process as well as boost the performance. Keywords: Digital watermarking, DWT-PCA, PSNR, Image denoising, convolutional neural network, DnCNN, residual learning, batch normalization.