Multi-scale signal-to-noise driven fusion of post-processing sequences for enhanced defect detectability in active infrared thermography

Active infrared thermography has emerged as a crucial tool in non-destructive testing, providing real-time visual representations of thermal patterns on material surfaces. However, detecting and analyzing defects can be challenging due to noise interference, the lack of standardization in post-processing techniques, complexity in data analysis, variability in defect visibility across frames, and the influence of environmental factors. To address these limitations, this study proposes a novel approach that enhances defect detectability by fusing multiple sequences derived from various post-processing methods into single, interpretable images. The proposed approach employs a multi-scale signal-to-noise ratio metric to accurately identify regions of interest and determine the optimal time at which defect detectability is maximized. Validation with two composite specimens featuring diverse defect characteristics demonstrates the capability of the method to simplify analysis and reliably improve detection performance. Compared with wavelet-based image fusion, the proposed approach achieves superior defect visibility and clarity, demonstrating a significant advancement in the effectiveness and reliability of thermographic inspections.