Active thermography has gained broad acceptance as a Nondestructive Test (NDT) method for numerous in-service and manufacturing applications in the aerospace, power generation and automotive industries. However, the diffusive nature of the heat conduction process renders imaging of subsurface structures susceptible to blurring and degradation of the signal with feature depth. Although this constraint is fundamental, significant improvements in blur reduction, depth sensitivity and detection of subtle features have been achieved. These improvements have been facilitated by a Thermal Signal Reconstruction method, based on a least squares polynomial fit of the logarithm of the time history of each pixel. The process separates temporal and spatial nonuniformity noise components in the image sequence, and significantly reduces temporal noise. Time derivative images created from the reconstructed data allow detection of subsurface defects at earlier times in the sequence than conventional contrast images, significantly reducing undesirable blurring effects, and facilitating detection of low thermal contrast features that may not be detectable in the original data sequence.