Digital Watermarking Algorithm Based on Neural Network in Multiwavelet Domain

A novel blind digital watermarking algorithm based on neural networks and multiwavelet transform is presented. The host image is decomposed through multiwavelet transform. There are four subblocks in the LL-level of the multiwavelet domain and these subblocks have many similarities. Watermark bits are added to low-frequency coefficients. Because of the learning and adaptive capabilities of neural networks, the trained neural networks almost exactly recover the watermark from the watermarked image. Experimental results demonstrate that the new algorithm is robust against a variety of attacks, especially, the watermark extraction does not require the original image.