The parameter estimation of frequency hopping signal in compressed domain needs to find the digital characteristics of compressed sampling data by means of measurement matrix, which results in high computational complexity and base mismatch. In order to solve this problem, a method for parameter estimation of frequency hopping signal based on digital characteristics in compressed domain and atomic norm is proposed. Firstly, a block diagonalization measurement matrix is established to realize signal piecewise compression, and the digital characteristics of compressed sampling data are analyzed to roughly estimate hop timing. Then, the signal segments without frequency hopping are separated and the accurate estimation of hopping frequency is realized by minimizing atomic norm. Based on the accurate estimation of hopping frequency, an atomic dictionary is designed and the accurate estimation of hop timing is realized in compressed domain. The proposed method's frequency estimation performance is better than that based on grid compressive sensing, and the hop timing of frequency hopping signals can also be accurately estimated. Simulation results show that when SNR is higher than -2 dB and the compression ratio is higher than 0.5, the normalized hopping frequency estimation error based on the proposed algorithm is less than 10-4, and the normalized hop timing estimation error is less than 10-2.