A Novel Watermarking Detector Based on Confidence Evaluation and Neural Network

Digital watermark is an effective copyright protection method. A new robust public watermarking algorithm is proposed with the base of the method of confidence evaluation and neural network data fusion. First a special addition formula is applied to later evaluate the confidence of each extracted watermarked bit in the extraction. Then a normalization is used to accord with the approximate distortion rate. Moreover the neural network is applied to extract the watermark. The algorithm doesn't need the original image during the extraction. The new watermark confidence evaluation relatively reduces the volume of the embedding watermark compared with the normal method. Finally the experiments show that the watermarking algorithm is imperceptive and more robust than the normal detector and average detector.