Abstract: To encrypt and compress the gray image simultaneously, a novel encryption scheme based on parallel compressive sensing and chaotic theory is proposed, which can form a balance between compression and security performance. Firstly, the wavelet coefficients of plain image are scrambled by Arnold permutation algorithm, and then compressed by the key-controlled partial random block weighing matrix. Finally, the cipher image is obtained after quantization and diffusion. Lorenz chaotic system is introduced to generate the key stream for permutation, construction of measurement matrix and diffusion. Via scrambling the sparse coefficients and constructing the measurement matrix with high reconstruction performance, the compression performance of scheme is enhanced. Moreover, the security and robustness are demonstrated by the security analyses.