Superconducting nanowire single-photon detectors (SNSPDs) play an important role in lidar and quantum information due to their excellent comprehensive performance. However, it is a major challenge to read out the SNSPD arrays for the high heat load and cost of the readout circuit in a cryocooler. Here, we propose a single-channel readout and imaging scheme based on compressed sensing (CS), which synergistically samples, combines, and reconstructs the response signals from all of the pixels of SNSPD arrays. Based on the row−column (RC) multiplexing SNSPD array, the compressed sampling of each pixel is realized by randomly switching the bias current of rows and columns, and the counts of each pixel are reconstructed according to the sampling matrices and combined readout signals. This scheme shifts the system complexity to the bias end, eliminates the need for expensive multichannel readout equipment, and enables multiphoton coincidence detection. We demonstrate spot imaging with a 16-pixel SNSPD array and find that the peak signal-to-noise ratio of the CS readout signal is 10 dB higher than that of the RC readout scheme at −100 dBm light intensity and 1000 fps.