Research On Digital Watermarking Based On Transform Domain And Probabilistic Neural Network

With the rapid development of internet and multimedia technology,a large amount of multimedia information crowd into the internet.The people are enjoying the convenient which is brought by the information.At the same time,many problems are along with,such as Illegal copying,illegal modification,unauthorized transmission and so on.The copyright protection problems of digital multimedia attract the attention from the researchers.Many protection schemes widely used,digital watermark is one of the most efficient means.The image regarded as the main part of the multimedia on the internet,its copyright protection is currently research focus.In this paper we review the history and current research status at home and abroad firstly.Because of its good performance of invisibility and robustness,transform domain digital watermark has been a hot area of research,which consists of DFT,DCT,DWT and CT.In this paper we introduce the intelligent algorithm in the field of digital watermarking.The detail is as follows:1)We propose a digital watermarking algorithm based on DFT and PNN.In the embedding processing,the host image is divided into blocks firstly.Then the blocks are decomposed by DFT,and we choose the amplitude image to embed the Pseudo Random Noise(PN).We use the relation between the location of the embedding and the PN to train the PNN.At last,we perform the inverse DFT to restore the host image.In the extraction processing,the first few steps are same with embedding.We use the trained PNN to determine the type of PN to restore the watermark which we embed into the host image.This algorithm enhances the invisibility and robustness.2)We propose a digital watermarking algorithm based on DCT and PNN.In the embedding processing,the host image is divided into blocks firstly.Then the blocks are decomposed by DCT.We choose the intermediate frequency to embed the PN.We use the relation between the location of the embedding and the PN to train the PNN.At last we perform the inverse DCT to restore the host image.In the extraction processing,the first few steps are same with embedding.We use the trained PNN to determine the type of PN to restore the watermark which we embed into the host image.This algorithm improves the ability of the invisibility and anti-compression.3)A digital watermarking algorithm is proposed based on DWT and PNN.In the embedding processing,the host image is divided into blocks firstly.We use the two-level DWT to deal with image.We choose the high frequency of the second level DWT to be the embedding location.We use the relation between the location of the embedding and the PN to train the PNN.At last,we perform the inverse DWT to restore the host image.In the extraction processing,the first few steps are same with embedding.We use the trained PNN to determine the type of PN to restore the watermark which we embed into the host image.The experiment results show that this algorithm gives a well performance.4)We propose a digital watermarking algorithm based on DWT and PNN.The host image is divided into several blocks,and then each block will be decomposed by CT firstly.Then coefficients of each block will be divided into many small coefficient blocks,the PN is embedded into the coefficient blocks that we selected with certain intensity.Thirdly,the correlation between each of embedded coefficient block and same-sized PN sequence is calculated,and put as the input of the probabilistic neural network system for training.Lastly,after training,the trained system will be robust for watermarking extraction.Results show that this scheme is blind,strongly robust and perceptual invisible.