The present work introduces a different data processing strategy, proposed in order to improve sub-surface defect detection on industrial composites; in addition, a resume of thermal data processing with most common algorithms in literature is presented and applied with new data. A deep comparison between the common absolute contrast, DAC, PCT, TSR and derivative methods and a new proposed contrast mapping procedure is implemented. Thermographic inspection was done in reflection mode on a Glass Fiber Reinforced Plastic plate, with flat bottom hole defects. Thermal data computation method is found to be critical for simultaneous defect detection and automatic mapping, optimized to identify defect boundaries at specific depth, with help of accurate image processing, implemented in a Matlab GUI for a reliable and rapid characterization of internal damage. The new processing approach, the Local Boundary Contrast method, elaborates different contrast maps and facilitates recognition of damage extension. Tanimoto criterion and the signal-to-noise ratio method were applied as a criterion to assess defect detectability of various processing methods.