Research Of Reversible Image Watermarking On Spatial Domain

With the growing popularity of computer technology and the rapid developmentof network based applications,the application range of the information technology(IT)is enlarging,people's requirement for security of the information is higher and higher,how to effectively and safely transmit and store digital information has become extremely urgent.However,the cryptology technology only protects the data during the data transmission from the sender to the receiver,As a suppelment and delvelopment,digital watermarking arises at the historic moment,solves the lack of cryptology,and has become a hotspot research in the field of information security,which provides a new thought and method for digital information communication safety and storage.Reversible watermarking is a new and important branch of digital watermarking,and is a technique to embed the watermark into digital media in a reversible way,the cover media can be restored intact after the embedded watermark and auxiliary data are extracted.An ideal reversible watermarking method should has the high embedding capacity and good quality of the watermarked image,the both properties impact and restrict each other,so the balance is the key for improving the reversible method.This paper analysises the reversible image watermarking method on spatial domain by utilizing the prediction algorithms,analysising histogram shifting technology,combining the characteristics of digital images.Our main research work and productions are listed as follows:(1)The existing histogram shifting based reversible watermarking methods do not have consider the shifting distortion,in order to solve this problem,we proposed a new reversible watermarking framework based on asymmetric error histogram shifting.And then based the new framework,we proposed a simple and efficient reversible watermarking method.The framework first selects a suitable predictor as the predictiton kernel,instead of the traditional algorithms,which calculate one prediction value for every pixel,named single prediction scheme.we design a multi-prediction scheme,which calculates multiple prediction values for each pixel to be processed,then define an asymmetric selection function,it will select the suitable prediction error,thus create an asymmetric prediction error histogram,which is different from the error histogram created by other reversible watermarking algorithms up to now,the method creates the histogram with the asymmetry pixel number distribution in the left and right sides,so during the data embedding,when we shift the fewer side of the histogram to create extra space for embedding the watermark,the shifting distortion that come from the shifting of histogram bins is reduced,which improve the quality of the watermarked image.Moreover,the traditional reversible methods usually shift the histogram bins to the left and right sides from the peak the peak and sub-peak points,respectively,thus obtain free space,and then choose the peak and sub-peak points to embed watermark for improving the embedding capacity.In the proposed method,we use the maximum and minimum functions respectively to construct the two error histograms,in which the maximum histogram is tend to the right and the minimum histogram is tend to the left,then embed watermark hierarchically by combining the two error histograms,since the two histograms are shifted in the opposite directions,which makes some watermarked pixels that have been modified in the max-embedding will be restored to their original values.Theoretical analysis and experimental results show that,comparison with similar algorithms,the proposed method increases the embedding capacity and reducs the shifting distortion,so it improves the quality of the watermark image,especially in the high embedding rates.(2)On the basis of the asymmetric error histogram with reducing shifting distortion,we proposed a reversible watermarking method based on directed-prediction scheme by modifying the high accuracy prediction algorithm-gradient adjusted predictor(GAP).In practice,since the prediction precision is high in the smooth region and the prediction precision is low in the complex region,so in order to reduce the image distortion,the data are embedded in the smooth region as more as possible,whereas the data are embedded in the smooth region as less as possible.Therefore,based on the GAP,this method increases the classifications in the smooth image region,so as to make the prediction accuracy be higher,and then design two directed-prediction scheme——RGAP and LGAP,the two schemes calculate prediction values in the opposite directions of the shifting in the complex region,thus it reduces the probability of the pixels by shifted in complex region and improves the quality of watermarked image.Finally,the optimal embedding points and thresholds are calculated by defining an evaluation index of the shifting distortion.Experimental results show that,in the predictiction results,the directed prediction scheme largely reduces the shifting distortion and improves the prediction accuracy,in the embedding process,the new formula of shifting distortion can find out the optimal embedding parameter,and can improve greatly the embedding effect,especially in the low embedding rates.(3)Prediction error expansion(PEE)is a technology that is researched widely in reversible watermarking method now,this technique can provide a high embedding capacity.In these methods,the key to improve the algorithm is how to reduce expansion distortion,which is caused by creating the extra space.Based on the analysis the characteristics of algorithm,by combining difference expansion(DE)and prediction error expansion(PEE),the first is to use the prediction errors pair instead of the pixels pair,then use the difference of the prediction errors pair replace the difference of pixels pair,this process has two advantages:1)Due to the introduction of prediction algorithm,prediction scheme with more high precision can be applied to it,so as to solve the trouble of the difference expansion,in which the prediction precision is low because of only the nearest prediction scheme-NNP can use in it,thus it reduces the expansion distortion of the transformation;2)Due to the existing of the correlation of image pixels,in most cases,the probability of error pairs with the same number is greater than the opposite sign,which will offset some parts of the errors,it reduces the expansion distortion further.So as to put forward a efficient reversible watermarking algorithm based on prediction error expansion.Theoretical analysis and experimental results show that,the proposed watermarking method improves the embedding capacity,and obtains higher quality of watermarked image.(4)Based on the idea of the reversible watermarking using the multi-bits error pair expansion(MMEPE),we first extend the integer transformation for 1-bit per pixel pair to multi-bits per pixel pair,then we designed a compound and adaptive embedding strategy,in small errors,the PEE technology is used to embedded for improving the embedding capacity,and in middle errors,the PEE technology is used to embedded for improving the quality of the watermarked image.Morover,in order to improving the precision precision,we proposed a novel error prediction algorithm,it includes four different types of predictors,which is 0,45,90 and 135 degrees prediction,respectively.In this algorithm,PEE and MEPE are predicted by the different prediction models,respectively,the former combines the probability prediction model,so as to gain high prediction precision;the latter combines the directed-prediction scheme,calculate asymmetric the prediction value,thus reduce the difference of the error pair,which increases the number of error pairs with the same sign,and reduces the expansion distortion,so it improves the quality of the watermarked image.Theoretical analysis and experimental results show that,the method overcomes the defects of DE that the watermark image quality is low,inherites the advantage of PEE that the embedding capacity is larger.This article provides a novel perspective and idea for the research of reversible watermarking technology in spatial domain,the research results can be more easily generalized to other digital media(such as audio,video),it provides the technique support for the watermark embedding of sensitive content in digital media,and can be used widely in information security field,such as copyright protection.