Multiple remote sensing image encryption scheme based on saliency extraction and magic cube circular motion

To improve the salient region security and transmission efficiency of remote sensing images, a multiple remote sensing image (MRSIs) encryption scheme based on saliency extraction and magic cube circular motion is proposed in this paper. The scheme provides two tiers of privacy protection for salient regions (airport locations) in remote sensing images. First, a 4D improved discrete tabu learning neuron (4D-IDTLN) chaotic system is proposed. The analysis of phase diagrams, the Lyapunov exponents (LEs) spectrum and bifurcation diagrams of the system show that the system exhibits rich dynamic behaviors. Second, the salient regions of remote sensing images are classified and extracted via knowledge-oriented saliency (KOS) and vision-oriented saliency (VOS) techniques to create a mask contour positioning model (MCPM) for the salient regions, which is then encrypted. Then, the MRSIs are fused into a cube, which is encrypted using magic cube circular motion and chaotic sequences and further encrypted using closed-loop diffusion. Finally, the security of the proposed encryption scheme is evaluated. The results indicate that the scheme provides higher security and better transmission efficiency for MRSIs.