Active infrared (IR) thermography is a non-contact, fast and wide-area nondestructive testing (NDT) technique that has been increasingly used in aerospace applications to detect both manufacturing and in-service environment-induced defects. The classical pulsed IR thermographic testing suffers from the problem of false indications due to surface clutter conditioned by uneven optical properties across a test sample surface. To some extent, this problem can be relaxed by using a technique called 'normalization'. A simple normalization algorithm, conventionally called 1D, involves the division of all images in a sequence by a chosen single image, often captured immediately after a flash. The novel technique of 3D normalization is implemented in this study to overcome the problems arising due to non-uniform heating and lateral heat diffusion in the case of pulsed IR thermographic NDT. In this case, normalization is carried out by dividing an experimental IR image sequence by the synthetic sequence which is calculated by solving the corresponding 3D problem of heat conduction. As a result of 3D normalization, the artefacts appearing due to uneven heating and absorption can be subdued through data normalization (division). Furthermore, fully automatic defect detection becomes possible as defining a reference point for sound area is not required as opposed to thermal contrast methods. In this work, the effectiveness of proposed 3D normalization approach is demonstrated for different heating conditions applied to glass and carbon fiber reinforced composites. (C) 2015 Elsevier Ltd. All rights reserved.