Compressive Sensing-Based HDR-Like Image Encryption and Artifact-Mitigated Reconstruction

High dynamic range (HDR) services can provide enhanced visual quality with brighter highlights and richer color details, which are appealing to consumers. However, the self-similarity of HDR files and the need for format conversion raise concerns about communication overload and security. While compressive sensing (CS) effectively compresses and encrypts images simultaneously, its protection remains weak for HDR-like images. Additionally, applying CS to HDR-like images can introduce color and motion artifacts during reconstruction. To address these challenges, a Compressive sensing-based HDR-like image Encryption and Reconstruction (CHER) method is proposed to achieve efficient compression and enhanced encryption while mitigating reconstruction artifacts. Specifically, CHER minimizes bandwidth requirements by compressing HDR-like images with CS and incorporating the four-dimensional Lorenz hyperchaotic system to avoid transmitting the CS measurement matrix. Moreover, a three-dimensional cube transformation is proposed and incorporated into the encryption process to enhance the confusion and diffusion effect of HDR content. Noticing the problems of two kinds of artifacts in the fused image, CHER takes advantage of cross-channel priors and cross-exposure priors for joint reconstruction. Experiments and analysis demonstrate that CHER can mitigate artifacts in an efficient and privacy-preserving way.