Experimental optical encryption based on random mask encoding and deep learning

Abstract

We present an experimental scheme for optical encryption using random mask encoding and deep learning technique. A phase image is encrypted into a speckle pattern by a random amplitude modulation in the optical transmission. Before decryption processing, a neural network model is used to learn the mapping relationship between the pure-phase object and the speckle image rather than characterizing the filter film used in the scheme explicitly or parametrically. The random binary mask is made by a polyethylene terephthalate film and 2500 object-speckle pairs are used for training. The experimental results demonstrate that the proposed scheme based on deep learning could be successfully used as a random binary mask encrypted image processor, which can quickly output the primary image with high quality from the cyphertext.