Abstract: The air-water interface is a strong reflector.Therefore, the downward reflection of acoustic waves at the water surface forms a well-known ghost effect in marine seismic waves.In marine surveys, ghost notches on the frequency spectrum reduce the frequency bandwidth and resolution of seismic records, distort waveforms, obscure the identification of events, and increase the difficulty of inversion and interpretation of seismic data.Deghosting seismic data is a longstanding challenge and has become appealing to researchers studying seismic imaging.Berkhout proposed a deghosting method based on echo-deblending, highlighting the similarities between deghosting and deblending, and suggested implementing a technique similar to deblending in simultaneous source seismic waves to separate ghost waves.In marine 3D data acquisition, undersampling seismic data reduces the effectiveness of the present deghosting methods.For 3D marine seismic data, the challenges of deghosting comprise two phases involving handling the wavefield undersampling issue, followed by actual ghost signal removal.To address the undersampling problem of 3D seismic data, Berkhout suggested interpolating seismic data before applying echo deblending to suppress ghosts.In this study, we proposed a 3D-deghosting algorithm for undersampling seismic data based on Berkhout's research.Our algorithm simultaneously addressed the undersampling and deghosting problems.The deghosting method for 3D seismic data based on densely sampled wavefield reconstruction did not use prior interpolation but extended the Berkhout method to establish a dense ghost wavefield on the water surface.It used Lasso regression to solve the problem of sparse extrapolation of the dense wavefield.Our algorithm first reconstructed the dense seismic wavefield at a specified surface by extrapolating the original sparse seismic wavefield.The dense ghost wave field at the water level was calculated using the dense wave fields and the echo-deblending approach.The sparse ghost wave field at the original acquisition position was then obtained.The ghost waves were subtracted from the original data using the least-squares method to resolve the deghosting problem.The goal was to minimize the difference between the actual data and the simulated wavefield, solve the uncertainty of this problem, achieve deghosting, and improve the characteristics of the seismic data.Synthetic data testing and real data processing showed that the deghosting of 3D seismic data based on dense field reconstruction could effectively separate ghost waves, improve wave field characteristics, and improve the resolution of the seismic data.