Reservoir computing based encryption-then-compression scheme of image achieving lossless compression

Abstract

Image encryption-then-compression (ETC), combining encryption and compression techniques, is a powerful strategy for image privacy protection. One of the most significant concerns in ETC is to realize a trade-off between high compression and high-quality recovery. To this end, we propose a novel ETC scheme based on a ring-network optical dynamic system and a reservoir computing (RC) system. This optical dynamics system is newly constructed by using three QD spin-VCSELs, then the optical RC system is further specifically designed. To ensure security, we develop a secret key generation method, which involves jointly injecting the original image and chaotic data into the RC system for chaotic transformation to produce high-security secret keys. In our scheme, the original image is first encrypted through scrambling and diffusion using the generated keys, then the obtained encrypted image is segmented into multiple sub-blocks, followed by using uniform down-sampling to select data, the unsampled data will be discarded and the sampled data is to be retained as the resulting compressed image. Flexible compression ratio can be achieved by varying sub-block sizes, with larger sub-blocks yielding higher compression. To realize high-quality recovery, the relationship between the compressed image and the discarded data is constructed in a learning manner by means of the designed RC system, which is conducive to the reconstruction process, resulting in lossless recovery. This work can effectively overcome the limitations of existing schemes in terms of simultaneously achieving both well-performing compression and lossless reconstruction. Experimental results are given which demonstrate the high security and superior performance of the proposed scheme.