Neural Network Oriented RONI Prediction for Embedding Process with Hex Code Encryption in DICOM Images

Advancements in digital era leads to several innovations which helps in sharing and remote accessing of clinical DICOM image format. While utilizing the updated facilities of health care innovations, one key issue is to confirm that the image falsification induced because of the watermarking procedure does not compromise the image analysis value. As watermarking process mainly focuses on the hiding of data into to the image in order to provide confidentiality and integrity to the sensitive information. Hence there is a need to recheck the quality of image after the process has been completed. For this case, the technique of reversible watermarking is emerged to confirm the quality of image at the end user side. This technique is a significant procedure in the applications which are requiring high quality image like medical sector. In this exploration effort, a novel image reversible watermarking is proposed called CNN oriented Region of Non-Interest (RONI) identification method in diagnostic imaging. This work is having two phases, 1) to implant a watermark in Region of the Non-Interest (RONI) and 2) for the adaptive derivation in quality factor using neural network system. The liver cancer image is taken as an input image in which the cancerous area is the Region of the Interest (ROI) and the remaining area forms the RONI. The performance metrics are Mean Square Error(MSE), Peak Signal to Noise Ratio(PSNR), Entropy(E) and Correlation Coefficient(CC). The outcome validates that the projected reversible watermarking procedure does not destroy the quality of image as the watermark is implanted only in region of non-interest and is resistive to threats.