Glass fiber-reinforced polymer (GFRP) composites are widely used in aerospace, aviation and other military and civilian fields. However, GFRP composites may have several defects due to the damage caused by various factors during the manufacturing processes and in deployment. Terahertz (THz) time-domain spectroscopy (TDS) imaging technology has the potential to become a powerful complement of the traditional nondestructive evaluation (NDE) methods for GFRP composites. During the imaging process, different parameters in time domain or frequency domain can be chosen to carry out the imaging. Targeting at different defects, the parameters which can effectively detect the defects are not the same. In this paper, the wavelet-based image fusion method is used to effectively combine multiple THz reflection images based on different parameters, and a new image which contains all the defects can be obtained. As shown in the experiments, due to the application of the wavelet-based image fusion method in the THz NDE, the defect information can be detected completely, which cannot be realized by single parameter imaging. It provides a new technical method for the post processing of the THz images of composite materials.