Abstract: Glass fiber reinforced polymer (GFRP) are widely applied in aerospace, transportation, etc., due to the advantages of corrosion resistance and high strength. However, defects such as delamination and bubbles formed in manufacturing call for non-destructive testing of such materials. This paper analyzed and addressed the drawback that the signal characteristics of debonding defects of GFRP with different depths are unobvious and put forward the continuous wavelet transform to enhance the terahertz feature methodology (CWT). Accordingly, the terahertz image obtained after continuous wavelet transform was objectively evaluated by calculating the image contrast. The gaus2 wavelet basis function was selected to perform defect imaging on the transformed signal. The peak value is increased by 4.5 times compared with the original, the image contrast of the terahertz defect imaging after continuous wavelet transform is increased by 1.3 times. Finally, the identification of the 50 μm debonding defect at the 5 mm position of the 6 mm GFRP is realized.