Shenhu area in South China Sea includes extensive collapse and diapir structures, forming high-angle faults and vertical fracture system, which functions as a fluid migration channel for gas hydrate formation. In order to improve the imaging precision of natural gas hydrate in this area, especially for fault and fracture structures, the present work propose a velocity stitching technique that accelerates effectively the convergence of the shallow seafloor, indicating seafloor horizon interpretation and the initial interval velocity for model building. In the depth domain, pre-stack depth migration and residual curvature are built into the model based on high-precision grid-tomography velocity inversion, after several rounds of tomographic iterations, as the residual velocity field converges gradually. Test results of the Shenhu area show that the imaging precision of the fault zone is obviously improved, the fracture structures appear more clearly, the wave group characteristics significantly change for the better and the signal-to-noise ratio and resolution are improved. These improvements provide the necessary basis for the new reservoir model and field drilling risk tips, help optimize the favorable drilling target, and are crucial for the natural gas resource potential evaluation.