The laser ultrasonic technique (LUT) enables the analysis of complex signals and the extraction of the features of defects. The time-domain features can be used to quantify various defects, but are negatively affected by the coupling efficiency, measurement noise, and surface roughness during LUT. This paper describes defect analysis methods based on the power spectral density and wavelet packet energy. Defects are characterized using the power of low-frequency components after wavelet decomposition and the wavelet packet energy, and then linear relationships are established for two defect types. The R 2 values of the linear fitting for the characteristic variables of the defect are greater than 0.96. The experimental results show that the proposed methods provide effective and quantitative characterization of the defect size from laser ultrasonic signals.