Carbon fiber-reinforced polymer (CFRP) is increasingly used by large industries because its characteristics are more favorable than those of traditional materials. However, due to mechanical stresses, or even manufacturing defects, CFRP tends to have internal defects, thus lowering the quality of the industrial processes in which it is used and compromising safety. A method for the early detection of internal defects in these materials, applicable to active pulsed thermography experiments, is presented herein. This method aims to select the most relevant components identified via the independent component analysis of sequences of thermographic images (thermograms) generated from the inspection of CFRP sheets to acquire a synthesis image that highlights the defects in the material. An application of this methodology to a CFRP sheet detects at least 20 of the 23 defects considered.