Ground-penetrating radar (GPR) is a nondestructive method using electromagnetic radiation to locate shallow geological subsurface features and underground utilities buried in the ground. It has become a valuable tool in several applications,1,2 such as archaeological explorations, environmental engineering, and geological problems. The effective imaging of buried objects is a key part of GPR, and the efficiency and resolution of the imaging results are the measure of the imaging algorithm.3 The theories of the present imaging methods are based on diffraction tomography (DT), reverse time migration (RTM), range migration (RM), and back projection (BP). The principle of the DT algorithm is based on the first-order Born approximation which assumes that the buried object of interest is a weak scatterer.4 A few additional assumptions are also invoked during the process of DT derivation to simplify and linearize the nonlinear electric field integral equation. These assumptions incur a trade-off to the reconstruction of the buried objects especially for the practical usage when noise is present in the collected field data. Taking advantage of the multiple reflections in the propagation medium, the RTM algorithm allows high-resolution focusing.5 However, the number of transmitting and receiving antennas must be more than the number of scatterers in the medium. The RM algorithm can work well only when the imaging scene can be modeled as a single homogeneous medium.6 When the GPR antennas and buried objects are distributed in different media, the imaging result of the RM algorithm will be blurred or possibly not focused at all. The standard two-dimensional (2-D) depth migrations7 can recover the location and shape of buried objects with arbitrary precision, depending on the accuracy of the velocity model used. The BP algorithm is one of the most practical imaging methods because of its convenience and robustness, particularly when the imaging scene can be modeled as layered media.8 Based on the aforementioned theories, some of the improvements and optimization imaging methods have been advanced to distinguish the shape of buried objects in GPR imaging. A modified split-step method9 was applied to extract structural information from a complex synthetic data set as accurately as possible, based on the standard 2-D depth migrations. Furthermore, a synthetic aperture radar technique10 was implemented for GPR image reconstruction, which can recover the shape of buried objects. However, the present imaging methods depend too much on the application environment or prior knowledge of the medium being imaged, which limits the popularization and application of GPR technology.