Abstract <p>Infrared thermal wave imaging, a pivotal technology in active infrared thermography, has found widespread application in the characterization of defects. However, the transverse thermal diffusion phenomenon leads to significant attenuation of the heat wave signal, severely restricting the sensitivity of this technique in detecting minute defects. This paper introduces an advanced method for compensating transverse thermal diffusion, referred to as the WI-ITHFS approach. The method utilizes wavelet decomposition to extract high-frequency details from the optical flow field, thereby capturing more precise information regarding heat flow dynamics. Subsequently, the gradients of these high-frequency coefficients are computed, and the coefficients of the adaptive smoothing term in optical flow estimation are dynamically adjusted via an enhanced Sigmoid mapping function, with the aim of improving the accuracy of the estimation. A feedback iteration strategy is also incorporated to further enhance both the stability and precision of the optical flow estimation. The performance of the proposed method is evaluated against the traditional Horn–Schunck optical flow method in terms of deformation error, energy values, and gradient consistency, with results demonstrating the superiority of the proposed approach. Finally, the heat flow compensation model is optimized using accurate optical flow estimation, integrated with a fusion strategy combining nonlocal steering kernels and background subtraction techniques. The experimental results demonstrate that the WI-ITHFS method significantly outperforms existing methods, especially in the detection of minute defects within the subsurface of aluminum alloy thin plates, offering a notably higher signal-to-noise ratio (SNR) in terms of contrast enhancement and defect detection accuracy.</p>