Using chaos to encrypt images with reconstruction through deep learning model for smart healthcare

In recent years, the potential for digital images to deliver important information carrier in healthcare domains mainly derives from the possibility of intelligent or interconnecting heterogeneous devices. Because of the high value of medical images, there is an urgent demand to protect their copyright and prevent leakage. Further, insufficient memory and bandwidth mean that storing and transmitting images is extremely costly. To solve this problem, an encryption -and -compression -based technique for medical images is proposed. First, a chaosbased encryption is used to encrypt the image, which is then compressed by downsampling using lossy compression. Meanwhile, a customised deep learning model is adopted to reconstruct the image, thereby reducing the memory and bandwidth costs and improving the security of medical images. The entropy, correlation coefficient, and NPCR/UACI score achieved by the proposed scheme is 7.999849, -0.00148/-0.00050/-0.000300, and 0.99585/0.33479, respectively at low time cost. The experimental results demonstrate that the proposed algorithm achieved superior performance compared with existing methods.