Research On Reversible Data Hiding Scheme With High Capacity And Signal To Noise Ratio

DH(Data hiding)has been widely used in authorization for digital media,copyright protection,passing military messages and privacy protection.However,the key problem in traditional DH methods is that the cover image cannot be fully restored after extracting the message from the stego image.As we know,for some special applications such as military,medical and legal applications,even small changes in cover image should not be ignored.Accordingly,lots of RDH(Reversible data hiding)methods appeared.RDH algorithm will not only extract embedded message correctly from stego image and restore the cover image reversibly.Recently,the RDH algorithms are classified into the four major categories: RDH algorithm based on difference expansion(DE),RDH algorithm based on histogram shifting(HS),RDH algorithm based on prediction error expansion(PEE)and RDH algorithm based on dual images and interpolated image.However,there are some problems in RDH algorithms for us to solve.Such as how to build some distortion models for stego images,how to construct sharper histogram,how to find the balance point between embedding capacity and the quality of the stego image,how to improve the robustness of the RDH algorithms and how to construct some new performance evaluation system.To solve the RDH problems,we proposed the measure to overcome noises and built some distortion models of the stego images.In the meantime,we analyzed the overflow/underflow issues and put forward the corresponding solution.Finally,some sharper difference histograms are found.In this paper,the major innovations are:(1)Reversible data hiding scheme based on significant-bit difference expansion was proposed.The original cover image can be recovered without any distortion after the hidden data have been extracted if the stego-image remains intact,on the other hand,the hidden data can be robust against unintentional changes applying to the stego-image,such as image compression and sometimes unavoidable addition of random noise which is below a certain level and does not change the content of an image.The SBDE scheme decomposes pixels in a cover image into two parts,that is,the higher significant bits(HSB)and the least significant bits(LSB),and calculates the HSB difference between adjacent pixels.Bits are embedded into HSBs by shifting the HSB difference value histogram bins.The shift and shift rule are fixed for all HSB difference values,and reversibility is achieved.Furthermore,owing to the separation of HSBs and LSBs,minor al-teration applying to the stego-image generated by non-malicious attacks such as JPEG compression,which will not change the HSB values as well as the HSB difference values,and robustness is achieved.(2)A high capacity reversible data hiding scheme based on right-left shift was proposed.Based on the characteristics of rectangle-prediction-error distribution,i.e.,bin 0 is the highest bin,the other peak bins are distributed on the left and right sides of the highest bin in a nearly symmetrical way and many zero bins locate on both sides of the peak bins.We proposed a high capacity RDH scheme based on right-left shift.The peak bins are first shifted towards right direction to leave some locations for embedding secret message bits,then,they are shifted towards left direction leave more locations for more data embedding,which can reduce the distortion due to right shift.By analyzing the necessary condition for the predictive error,the issue of overflow/underflow can be easily solved without adding side information in the embedding data.Experimental results on different images illustrate the proposed scheme is of high embedding capacity(EC)and visual quality on different images,meanwhile,the theory demonstrates the performance of the proposed method outperforms other RDH methods.(3)A reversible data hiding algorithm based on bidirectional difference expansion was proposed.Scanned in a Z-shaped way,the cover image was first transformed into a one dimensional array.Then,the difference between two adjacent pixels was expanded towards two directions,and a bit of secret message was embedded into the left pixel.Finally,the array was transformed into a two dimensional matrix.After receiving the stego image,the stego image scanned in a Z-shaped way was first transformed into a one dimensional array.Then,the difference between two adjacent pixels was compressed towards two directions,and a bit of secret message was extracted from the left pixel.Finally,the stego array was transformed into a two dimensional matrix,so,the lossless cover image was obtained.Besides,to improve the quality of the stego image,we used the range of the mean of the two embedding pixels to solve the issue of overflow/underflow.Experimental results show,the EC of the cover image,the quality of the stego image and the security of the secret data are improved obviously.