Adaptive embedding: A novel meaningful image encryption scheme based on parallel compressive sensing and slant transform

A novel image encryption and adaptive embedding algorithm is proposed in this paper by combining 4D memristive hyperchaos, parallel compressive sensing (PCS) and slant transform (ST). This work dedicates to balancing the performance of existing visually meaningful image encryption algorithms which embed cipher image into host image in spatial or transform domains. First, the 2D discrete wavelet transform (DWT) is adopted to sparse the plain image. Then the sparse matrix after threshold processing is encrypted and compressed through Arnold scrambling and parallel compressive sensing to obtain the cipher data. Next the noise visibility function (NVF) is utilized to find the sub-images suitable for embedding in host image. After that the unquantified cipher data are randomly embedded in them by ST-based embedding method with block-wise manner and the final visually meaningful cipher image is obtained. To withstand the chosen-plaintext attacks (CPA) and the known-plaintext attacks (KPA), the frequency domain information of plain image is used to control the 4D memristive hyperchaos to generate the scrambling matrix. Additionally, the experimental results and comprehensive analyses demonstrate that compared with the existing related algorithms, the proposed encryption algorithm has good visual security, decryption quality and superior robustness. (c) 2021 Elsevier B.V. All rights reserved.

Automated summary

This paper proposes a novel image encryption and embedding algorithm by combining multiple techniques to achieve visually secure image encryption. By optimizing the processing of sparse matrices and embedding methods, high-quality cipher images are obtained.