We developed a novel method to improve the computational efficiency of full-waveform inversion (FWI). FWI has drawn more and more attention in many scientific and engineering fields because quantitative information and high-resolved imaging can be obtained. One of the shortcomings of FWI is its high computational costs both the computational memory and calculation time. We tackle this problem by introducing an adaptive refinement technique of the spatial resolution. The method is based on a mesh-free finite difference method, which is one of the true mesh-free methods. In our strategy, the spatial resolution is refined based on a velocity model in a simple manner, i.e. a calculation point located in a low velocity zone is divided into smaller ones to provide higher resolution. The procedure is conducted during nonlinear inversion analysis adaptively. This strategy can avoid the numerical dispersion without additional complex processes, for example, the mesh generation procedure in finiteelement methods. We examine the effectiveness and efficiency of the proposed method using numerical experiments. Our result shows that the method can obtain the accurate velocity structure successfully with the local refinement technique. Our inversion result shows good agreement with that of the conventional finite difference method. We also compare the computational costs of the method with those of the conventional finite-difference method. The numerical experiments indicate that the proposed method can be an alternative to the conventional methods.