Thermal non-destructive testing (TNDT) is one of the emerging inspection and evaluation techniques mostly used for subsurface defect detection in various industrial components. Besides the conventional thermography techniques (such as lock-in and pulse), recently introduced non-stationary thermal wave imaging (NSTWI) techniques gained its applicability in TNDT community due to their inherent testing capabilities such as improved sensitivity and enhanced resolution in inspecting and evaluating various solid materials for detecting subsurface defects. Barker-coded thermal wave imaging (BCTWI) is a one of the widely used NSTWI techniques, which facilitates the use of low peak power heat sources in moderate experimentation time in contrast to conventional TNDT techniques. In this paper, the pulse compression favorable NSTWI (BCTWI), the reconstructed pulsed (main lobe) data have been considered and processed using independent component analysis and named Barker-coded independent component thermography (BCICT). This proposed BCICT is implemented on a mild-steel sample to detect the artificially simulated flat bottom circular holes located at different depths inside it. The proposed technique extracts the sub-surface details such as flaws/defects hidden inside the sample by an unsupervised learning process, which helps in eliminating the manual interpretation of subsurface defects. The applicability of the proposed algorithm has been evaluated and validated experimentally with two different excitations schemes by considering the contrast and signal-to-noise ratio (SNR) as figure of merit. The results indicate that the BCICT technique offers higher contrast and SNR in comparison to conventional pulse-based TNDT technique.