Triple Color Image Encryption Using Hybrid Digital/Optical Scheme Supported by High-Order Chaos

Abstract

Image security has attracted increasing interest, wherein image encryption is an effective and direct method that can be implemented using digital, optical, or hybrid techniques. The challenge is how to design a high-speed, and high-security level multiple color image encryption scheme which makes use of advanced techniques such as chaotic system and deep learning. This issue is addressed in this paper where a nine-dimensional (9D) chaotic-based Hybrid Digital/Optical Encryption (HDOE) scheme is proposed for triple color images. The scheme consists of cascading digital and optical encryption parts controlled separately by the chaotic sequences. The nine chaotic sequences are grouped into three sets, and each set is responsible for the encryption of one of the RGB channels independently. The digital part uses fusion, XOR operation, and scrambling. The optical part uses two independent chaotic phase masks in the optical Fourier transforms domain. A Denoising Convolution Neural Network (DnCNN) is designed to assist the robustness of the decrypted images against Gaussian noise. The simulation results reveal that the proposed triple-image HDOE scheme offers entropy of 7.9991, 7.9987, and 7.9991 bits for R, G, and B channels, respectively,  and infinite Peak Signal-to-Noise Ratio (PSNR) for the decrypted images.