Research On Reversible Data Hiding For JPEG Images

“Internet + transportation” brings great convenience to people’s lives,but also brings the problems regarding information security.The most common thing in the transportation system is the exchange of digital multimedia information,where third parties are allowed to access all types of multimedia information.In such cases,when users participate in the intelligent transportation system,the information of passengers may be leaked.Personally identifiable information has a very bad impact on the lives of passengers,and even brings huge economic losses to related companies and departments.Therefore,needing effective technologies to enhance the security of Internet information in the transportation system urgently.It is well known that reversible data hiding(RDH)is an efficient technique to solve the above problems.RDH has the advantages of not only extracting the embedded data correctly,but also recovering the original cover image without any distortion.Applying RDH to the transportation system can distinguish the authenticity of electronic tickets,maintain the copyright,and protect the personal information of passengers.Joint Photographic Experts Group(JPEG)is the most widely used compression format in daily life due to its high compression ratio and satisfactory visual quality.Considering that the redundancies of JPEG images are eliminated largely,RDH for JPEG images inevitably leads to low embedding capacity.In addition,embedding the secret data into the cover images also cause the file size increment of JPEG images.Therefore,under a given embedding capacity,how to design an RDH algorithm to reduce the embedding distortion and keep low image file size increment is a hot and difficult research area.In this paper,on the basis of ensuring the reversibility and pursuing high embedding performance,the following researches have been carried out to increase the embedding performance while maintain low file size increment:Considering that the one-dimensional histogram cannot make full use of the correlation between adjacent pixels,an adaptive RDH method for JPEG images based on multiple two-dimensional(2D)histograms is proposed.In this method,multiple 2D histograms are constructed by pairing two adjacent non-zero coefficients in the middle and low frequency bands,and four 2D mapping are designed for each 2D histogram.According to the desired embedding capacity and the local characteristics of each 2D histogram,the 2D mapping with the best rate-distortion performance is adaptively selected for each 2D histogram,which greatly reduces the embedding distortion and obtaining low file.To reduce the embedding distortion as well as high computational complexity of multi-histogram RDH while maintain low file size growth,an efficient adaptive RDH method based on multiple 2D histograms for JPEG images is proposed.Smoothness thresholds are adaptively selected according to the local properties of histograms to construct multiple sharply-distributed 2D histograms.Next,coefficient pairs of each histogram in smooth regions are preferentially selected to generate the corresponding 2D histogram.In the following,the optimal mapping of each 2D histogram is adaptively selected to achieve higher rate-distortion performance.Finally,to reduce high computational cost,IDPSO is used to solve the optimization problem.For the proposed methods above,different datasets are used as test datasets in the experiments.This article compare the embedding performance of a single image and the average embedding performance of multiple images in different datasets with several related works.Besides,the image visual quality,the file size change before and after embedding are evaluated.Subsequently,the feasibility,advantages and disadvantages of the algorithm are analyzed and summarized,and the direction to be solved in the next research is given.