Optical color image cryptosystem based on interference principle and deep learning

Abstract

In this work, we propose a method for optical color image encryption based on interference principle and deep learning technique. After the interferograms being collected through three channels, a color image to be encoded can be captured optically and transformed into a color interferogram. An end-to-end DL model is trained to derive the mapping relationship between the interferograms and their original plaintexts. In this method, Arnold transform is used to scramble the recorded interferogram for achieving high security. The primary color image can be decrypted by using inverse Arnold transform and the trained learning model. The method provides the ability to decrypt a color image from only one color interferogram, which couldn’t be achieved by using conventional methods. In the decryption process, the optical keys used for traditional decryption are not required. The effectiveness and security of the cryptosystem are verified by computer simulations.