Abstract: There are a substantial amount of gas-bearing sediments occurred in the Qiongdongnan area. Geophysical features, such as in-phase axial pulling down and velocity changes are commonly observed. Therefore, the images of sediments below the gas hydrate layer are always blurring. In this paper, a method based on non-linear tomographic depth offset is adopted to enhance the accuracy and resolution of imaging. The method of full 3D tomography inversion is used to establish the velocity model in depth domain, and the kinematic invariants are calculated by picking up the RMO of depth offset gathers and reverse deflection. Based on the mixed model under the constraints of skeleton information, such as horizons and inclination angles, kinematic invariants are used for velocity tomography to minimize RMO for model updating. This method can avoid multiple iteration PSDM processing while updating velocity and greatly improve the tomographic efficiency. Then the high precision velocity model can be obtained by making full use of the profile skeleton and inclination information. The effect of gas on stratigraphic formation is effectively eliminated in the practical application of hydrate data in the Qiongdongnan area, and high precision depth imaging results are obtained.