With the rapid development of computer science, the storage and dissemination of information are often carried out between various types of computer hardwares and various networks. The traditional information encryption scheme has gradually disappeared. Therefore, computer-based information encryption algorithms have gradually become a research hotspot in recent years. By combining the theory of wavelet packet transform, compressed sensing and chaotic system, a multi-process image encryption scheme based on compressed sensing and multi-dimensional chaotic system is proposed. The encryption scheme implements compression and encryption for grayscale images and corresponding decompression and decryption process. The wavelet packet transform theory is applied to the image preprocessing stage to perform wavelet packet decomposition on the original image. At the same time, the image signal components obtained by the decomposition are classified according to the threshold processing method, and the characteristics of the image signal components are processed in the subsequent processing. They are compressed, encrypted, or reserved in a differentiated manner. In the image compression stage, by introducing the compressed sensing algorithm to overcome the shortcomings of the traditional Nyquist sampling theorem, such as high sampling cost and low reconstruction quality, the compression efficiency and compression quality are improved while the ciphertext image reconstruction quality is guaranteed. In the image encryption stage, the encryption scheme combines multi-class and multi-dimensional chaotic systems to confuse and scramble the related image signal components, and introduces a high-dimensional chaotic system to make the encryption scheme have a large enough key space to further enhance the ciphertext image reliability. Finally, the complete reconstruction of the original image is achieved by applying the inverse of compression, encryption and wavelet packet transform. The simulation results show that the image encryption scheme effectively protects the basic information about ciphertext images by virtue of algorithm robustness against external interference, and does not reveal any useful information when dealing with cracking methods such as plaintext attacks. In addition, the information entropy and correlation coefficient of ciphertext images encrypted by this encryption scheme are closer to ideal values than those of the encryption algorithm in the references, and its encryption performance is significantly improved.