Research On High-performance Reversible Information Hiding Algorithm For Medical Images

The rapid development of multimedia technology leads to the creeping entry of electronic products into the medical field,which gives birth to the electronic medical system and brings the medical industry into the electronic age.Among them,telemedicine has become an important achievement in the development of electronic medical system,which breaks the spatial distance between patients and doctors.Electronic medical records are introduced as a result,a successful advance for the quality and efficiency of telemedicine.At the same time,problems of information security also exist in the transmission process of telemedicine,such as unauthorized access,disclosure of electronic medical record data and tampering of medical images.However,because of the embedding of secret information in the loaded image,there will be some distortion,and the greater the distortion,the easier it is to attract the attention of the attacker.In order to realize secure transmission of medical images and electronic medical records in telemedicine,reversible data hiding(RDH)technology plays an indispensable role in reversible data Hiding.In this technology,electronic medical records containing patients’ personal information and clinical reports are embedded into patients medical images as secret information,so as to obtain confidential medical images that do not attract the attention of attackers and make them transmitted in the medical system.In order to protect patients’ privacy and improve the visual security of loaded medical images during telemedicine transmission,a high-fidelity RDH framework is designed to hide and retrieve electronic medical records in medical images,which provides a new security mechanism for telemedicine data.(1)Firstly,a reversible data hiding algorithm based on reducing the number of pixel shifts in the histogram is proposed.By considering the regional characteristics of medical images,including the region of interest(ROI)and the region of non-interest(RONI),we combine a new regional division method and secret information mapping method to design an RDH algorithm based on reducing the number of pixel movements in histogram based on medical images.The aim is to reduce the number of pixel movement caused by distortion in the process of histogram movement,so as to achieve high fidelity of loaded medical images and reduce the attention of attackers.At the same time,a new prediction method is designed to generate a more suitable prediction error histogram(PEH)for embedding processes,so as to further improve the performance of the scheme.In this scheme,a direction predictor is first used to preprocess a given medical image,and different prediction schemes are used to calculate the prediction errors in each direction,among which the smallest prediction error is taken as the final prediction error to generate a single PEH.The mapping is then used to change the number of "1" in the binary secret message that cause a valid move.Finally,the embedding mechanism of PEH is improved to limit the movement range of invalid moving pixels.Then,according to the pixel sequence of RONI and ROI,the secret information after processing is embedded successively.(2)In addition,we have proposed another reversible information hiding algorithm based on asymmetric residual histogram movement to further improve the algorithm’s high fidelity performance.This algorithm designs the histogram generation process in the histogram moving algorithm by utilizing more correlation between surrounding pixels.Different from the previous algorithm to generate a single symmetric histogram,this algorithm generates two asymmetric histograms and uses the two histograms to embed the image pixels twice respectively.Through the complementary effect,that is,the movement direction of the positive and negative histogram embedding is opposite,partial watermark distortion can be recovered.At the same time,combined with the improved information mapping method,further reduce the distortion caused by effective moving pixels,and achieve high fidelity of the loaded images.Experimental results show that both algorithms have better visual quality compared with existing algorithms.Meanwhile,when embedding information with the same capacity of10000 bits,the second algorithm in this thesis increases its peak signal to noise ratio(PSNR)by about 1.6d B compared with the first algorithm.