Ultra-wideband frequency-stepped signal is an effective radar signal, but its anti-jiamming ability is not enough, and its pulse-width time is too long. To solve these problems, an imaging algorithm based on sparse frequency-stepped signal is proposed, the use of which can complete the target imaging with less resource of time and spectrum. In the proposed algorithm, the echo data of sparse frequency-stepped signals are equivalent to the observed values of the echo data of the traditional frequency-stepped signals. Therefore, the reconstruction of the high-resolution range profile can be realized based on the compressive sensing (CS) theory. Then, the target scene imaging can be completed by range cell migration correction and azimuth pulse compression. The simulation results demonstrate that the use of the algorithm can achieve a high-resolution imaging while the transmitted frequency-stepped signal is sparse in frequency domain. Finally, the ground-based radar experiment data are adopted to verify the effectiveness of the proposed algorithm.