A self-adaptive scheme for double color-image encryption

Most of existing optical color image encryption schemes have born security risks due to the adoption of linear transform, and data redundancy for the generation of complex image. To settle these problems effectively, a self-adaptive scheme for double color-image encryption is proposed in this paper. In this scheme, each RGB color component of two secret color images is first compressed and encrypted by 2D compressive sensing (CS) in which measurement matrices are generated by compound chaotic systems. Then, the two measured images are regarded as the real part and imaginary part, constituting a complex image to reduce data redundancy caused by following optical encryption. In the end, the complex image is reencrypted by self-adaptive random phase encoding and discrete fractional random transform (DFrRT) to obtain the final encrypted data. In the process of DFrRT and random phase encoding, the correlations between R, G, B components are adequately utilized. The production of key streams not only depends on the initial value but also on plain-text, and the three color components affect each other to enhance the ability against the known plaintext attack. The projection neural network algorithm is adopted to obtain the decryption images. Simulation results also verify the validity and security of the proposed method.