A multi-image encryption scheme based on block compressive sensing and nonlinear bifurcation diffusion

With the frequent cyberattacks and the rapid growth of image data volume, compressive sensing based image encryption algorithms have been widely studied. However, the existing schemes are usually imperfect in terms of encryption capacity, reconstruction quality and security. To address these issues, a novel multi-image encryption scheme is presented based on block compressive sensing and nonlinear bifurcation diffusion. Firstly, a block-level adaptive thresholding sparsification algorithm is devised to optimize the image sparsity and the decryption quality by dynamically adjusting the local thresholds for different wavelet coefficient regions according to their specific features. Secondly, to destroy the inter-block correlations while reduce the transmission burden, the orthogonal Hadamard matrix and the chaotic system are employed to construct a distinct measurement matrix for each image block. Finally, a nonlinear bifurcation diffusion algorithm based on the complete binary tree is developed to enhance the encryption performance through mutual diffusion among non-adjacent pixels. Additionally, the secure hash algorithm is employed to establish a close relation between plaintext images and secret keys to defy the chosen-plaintext attack. Simulation experiments and performance analyses demonstrate the feasibility and the reliability of our presented multi-image encryption scheme. Compared with some newly advanced schemes, our presented multi-image encryption scheme exhibits certain advantages concerning compression performance, key space, and resistance against statistical and differential attacks.