Gaussian beam migration (GBM) is a ray-based seismic imaging method,which can handle multi-arrivals and has good imaging accuracy.However,two problems should be taken into consideration when this method is used in complex topographic conditions:First,the original local plane wave decomposition method has insufficient calculation accuracy,which affects the imaging quality of low signal-noise ratio data.Second,the ray beam propagator is the most important factor for migration accuracy.With the enhancement of the propagation distance,the Gaussian beam width increases rapidly and cannot ensure the imaging accuracy of the near surface and depth structures at the same time.In addition,the insufficient coverage of the Gaussian beam in the shallow part of the model might affect the imaging quality of this region.In order to solve the above problem,the authors apply wave decomposition technique based on compressed sensing theory to the complex surface in Fresnel beam migration.It not only effectively improves the precision of local plane wave decomposition,but also solves the problem that the width of Gaussian beam operator is increased quickly with the increase of propagation distance according to the Fresnel beam operator.Typical numerical examples prove the validity and stability of this method.