From local to global analysis of defect detectability in infrared non-destructive testing

Several image processing techniques are employed in Infrared Non-Destructive Testing (IRNDT) to enhance defect detectability. To date, there is no adequate global measurement that objectively assesses defect visibility in processed frames. In this work, a Global Signal to Noise Ratio (GSNR) that comprehensively evaluates defect detectability in processed infrared (IR) images is proposed, as well as a defect visibility measure named Infrared Image Quality Index (IRIQI) that compares the structural information of defective and sound areas. In addition, GSNR and IRIQI are validated by using the area under ROC curve (AUC). AUC quantitatively assesses defect visibility by comparing the outcomes of processing techniques to human judgements. The remarkable benefit of this global approach is that it allows one to determine the frame at which processing techniques reveals the majority of the defects by evaluating the times at which AUC curves reach their maxima. The test pieces were a Carbon-Fiber Reinforced Plastic (CFRP) sample containing delaminations and a honeycomb specimen with delaminations, skin unbonds, excessive adhesive, and crushed core.