Reversible Watermarking Based On Adaptive Prediction Error Expansion

In the past two decades,digital watermarking technology has been widely used in the copyright protection and content authentication of multimedia.But,after the watermark information is extracted,the original image cannot be completely recovered.In industries dealing with highly sensitive data such as military,medical,and legal industries,even minor data distortions are difficult to be tolerated.Reversible watermarking can not only extract the watermark information at the receiver side,but also recover the original image without any distortion.However,there are still some problems existing in current algorithms:(1)In the traditional reversible watermarking algorithm based on prediction error expansion,the watermark bits are embedded into the two peaks in the global histogram of the prediction error.This scheme ignores some bins in local histogram;(2)Most reversible watermarking algorithms are designed for gray scale image.The research for color image is insufficient.To solve the first problem,a reversible watermarking based on locally adaptive prediction error expansion is proposed.The original image is divided into two regions.In the first region,the image is divided into small blocks and the local histogram of prediction error is obtained from checkerboard prediction.Then,the two peaks in the local histogram are used for watermark embedding.Since the collection of all peaks pairs in local histogram correspond to more bins in global histogram,the histogram can be fully explored to obtain more embedding positions.The second region is used to embed the auxiliary information and compressed location map.This scheme makes full use of the local prediction error.When embedding the maximum capacity,the peak signal to noise ratio(PSNR)of the test image is larger than 50dB.To solve the second problem,a color image reversible watermarking based on adaptive prediction error expansion is proposed.The algorithm uses the correlation between the three channels of the color image.The red channel or the green channel is used to predict to attain the local prediction error.The two peaks in the local histogram are used for watermark embedding in the blue channel.This scheme does not need to retain the original pixels used for prediction.In addition,in the process of embedding a color image,there is no need to save the expansion bin of each block.So,the capacity of one color channel is larger than that of the grayscale image.When the embedding rate achieves the embedding capacity,the PSNR of the watermarked image is larger than 47dB.The experimental results show that,in the improved grayscale image algorithm,the embedding capacity is improved for the same embedded image quality.In the improved color image algorithm,the available local prediction error used for expansion is increased.The embedded auxiliary information is reduced.So,the embedded capacity is improved.