In order to increase the detection capability of infrared modulated thermography for defects in composite materials, an infrared image sequence processing method using the complex Morlet wavelet phase analysis was studied. According to the modulation frequency of thermal excitation and the sampling frequency of infrared imaging, the total number of scales used in wavelet transform was selected adaptively, and then a phase image sequence was reconstructed from the raw thermal image sequence by using the complex Morlet wavelet coefficients. An experiment of modulated thermography was performed on a glass fiber reinforced plastic reference sample containing delamination defects, the application effects of the method were quantitatively evaluated and compared with the Fourier transform method by the signal-to-noise ratio. The results show that the complex Morlet wavelet phase method can separate the low frequency and high frequency noise from the thermal image signal, and extract the subsurface defects at different depth robustly; it keeps a good tolerance to the change of the modulation frequency; so it is an effective algorithm of the modulated thermography.