The high productivity blended acquisition technique greatly improves the acquisition efficiency. However, this technology also introduces serious interference noise from adjacent sources. An efficient data deblending method based on sparse inversion is used to separate the simultaneous sources for offshore 3D DAS VSP data. Due to the difference in coherence between main source and blended noise on the time-space common receiver point gathers, which can be expressed by the sparsity in the FKK domain, we can solve the effective signal by imposing additional constraints on the unknown signal model. The signal model with gradually improved signal-to-noise ratio can be obtained by subtracting interference noise from simultaneous sources. The method has been proved to be efficient and useful in separating the main source from the interference noise by using the Offshore 3d DAS VSP data. It should be pointed out that this method is not applicable to the data acquired in the area with irregular spatial distribution of shot points because this method requires 3D FKK transformation of common receiver point data.