<p>Recently, image encryption methods using compressed sensing (CS) technology for visual security have gained widespread attention. These schemes simultaneously encrypt and compress the plain image at one time, and then embed the encrypted image into a new carrier image (i.e., cover image), making the secret cipher image better be concealed in a visually secure manner. However, considering reconstruction quality and embedding efficiency, new hiding algorithm should still be studied further for improvement. To address these problems, a visual image hiding algorithm is presented in this paper, i.e., ViIHA. First, a pseudo-random sequence generation method using restricted Boltzmann machine is designed to improve the randomness of keystream. Second, a measurement matrix and a parallel CS technique are constructed to enhance execution efficiency. Next, a random dynamic embedding strategy is proposed to improve data embedding capacity while ensuring the visual quality for cover image. Finally, simulation experiments show that ViIHA has high security and strong robustness. Compared with some current algorithms, the proposed ViIHA can achieve a better quality for both reconstructed image and cover image, demonstrating a good performance.</p>