Imperceptible Image Watermark Embedding using Multi-resolution Wavelet Decomposition and Deep Convolutional Neural Networks
Amal S. Khalifa · 2023
Due to the ease of access to and unauthorized copying of multimedia data, especially images, digital watermarking techniques have become a necessity. In this research, we propose a method that combines the power of deep learning, and the multiresolution property of wavelet transforms for the purpose of practical and imperceptible digital image watermarking. The proposed method utilizes a pre-trained CNN model to fuse the watermark image into the approximation wavelet sub-band of the host image. The effectiveness of the proposed method is demonstrated through several experiments at different levels of wavelet decomposition. In addition, the method was tested under common third-party attacks such as compression and filtering. When compared with some existing techniques, experimental results showed the superior performance of the proposed method in terms of imperceptibility and hiding capacity.