We have developed a new Gaussian beam reconstruction algorithm using time-domain Gaussian beam (TGB) method to decompose seismic data. The TGB is characterized by a particular arrival time, location, amplitude, orientation, curvature, and extent. TGB decomposition and reconstruction of seismic data are implemented by the plane-wave decomposition (PWD) theory. First, we evaluate the construction principle of TGB, and then we develop the PWD filter to decompose seismic data into local plane waves by estimated dip fields and curvature fields of the seismic records. Next, the local plane waves in terms of TGBs are used to reconstruct seismic data through iteratively minimizing the residual error. Afterward, Gaussian beam depth migration is performed on the reconstructed data. Finally, we analyze the reconstruction results under the circumstance of seismic data with randomly missing traces. Numerical tests indicate that for data with missing traces, the Gaussian beam method obtains better reconstruction performance than the traditional projection onto convex sets method with the same number of iterations. The combination of Gaussian beam seismic data reconstruction and migration extends the research field of Gaussian beam migration, which has an important theoretical and practical significance.