A reflectance-correction retinex framework for thermal image enhancement in nondestructive defect detection of CFRP.

To ensure the quality of materials, pulsed thermography (PT) is an effective technique to detect the presence of internal defects in carbon fiber reinforced polymers (CFRP). However, non-uniform thermal excitation during the testing often leads to uneven heat distribution within the original thermal images, and the limited heat absorption by the surface of specimen may cause low-contrast defect signals that further impair the identification of defects. In response to these challenges, this study introduces an enhancement framework, termed Reflectance-Correction Retinex (RCR). In the RCR framework, the single-scale Retinex method is first applied to equalize the non-uniform background within a thermal image, and a sigmoid modification is proposed to adaptively amplify the contrast between defects and their background. The RCR framework is tested on PT images of man-made CFRP specimens. The experimental results show that our proposed method substantially enhances the visual quality of PT images, which facilitates precision and reliability in defect detection applications. [Display omitted] • Pulsed thermography is applied to detect internal defects inside CFRP. • An RCR framework is proposed to enhance visual quality of PT images. • SSR is applied to equalize non-uniform background of PT images. • An adaptive sigmoid modification is proposed for signal enhancement. • The RCR framework facilitates the accuracy of defect detection. [ABSTRACT FROM AUTHOR]