Abstract: Noise suppression of marine gravity data is one of the key steps to obtain real gravity field signals and improve the accuracy of gravity field data. At present, commonly used noise suppression methods suffer from problems such as time delay and residual dynamic noise. Therefore, we applied the minimum curvature method to the noise suppression of marine shipborne gravity measurement data. First, the measured gravity data were tested and analyzed. Secondly, the applicability and effectiveness of this method were verified. Thirdly, the noise suppression performance of the method was evaluated by testing different iteration numbers and iteration step sizes. Finally, a stepwise minimum curvature noise suppression method was proposed. The new method not only improved noise suppression but also mitigated the endpoint distortion phenomenon, and showed advantages of having stable computational results, no time delay, and flexible selection of calculation parameters. The external coincidence accuracy between the results of the method and those of FIR low-pass filtering reached up to ±0.131 × 10⁻⁵ m/s², demonstrating strong potential for application and promotion in future.