Digital Watermarking Based on Neural Network and Image Features

A novel watermarking technique based on iamge features and neural network is proposed. The watermark is the fusion of a binary copyright symbol and image feature label which is gotten by analyzing the image fractal dimension. Then the watermark is pretreated by Arnold transform in order to increase the security, and embedded into the multiwavelet transform domain. At the same time, the back propagation neural network is applied to improve its imperceptibility and robustness. Experimental results demonstrate that the proposed algorithm can achieve good perceptual invisibility and security, and it is also robust against various common image processing attacks.