The invention discloses a remote-sensing image large-magnification compression method based on a lightweight deep convolutional network. The method comprises the following steps: 1, constructing the lightweight deep convolutional network with seven layers; 2, selecting a remote-sensing image training sample; 3, generating a training data set; 4, training the lightweight deep convolutional network;5, compressing the to-be-tested remote-sensing image; and 6, decompressing a large-magnification compressed file of the to-be-tested remote-sensing image. A new lightweight deep convolutional networkwith seven layers is designed through the method disclosed by the invention; the model of the method is light and handy, the compression and decompression time are short, the large magnification compression is realized, the inverse quantization is optimized, the precision of the neural network is improved, the restored remote-sensing image quality is improved, and the edge texture information ofmore remote-sensing image can be reserved.