This study presents a comparative analysis of the novel guided-wave-based imaging method that integrates variational Bayesian principal component analysis with time-delay strategy for detecting internal and external defects in plate-like structures. The performance of the conventional delay-and-sum imaging method deteriorates when the signal-to-noise ratio of signals is low or when other wave packets overlap with the defect scattering signal. The imaging method based on variational Bayesian principal component analysis analyzes the principal components and corresponding singular values of the time-delayed signal array, and the maximum singular value represents the contribution of the most principal component, serving as an indicator of the coherent defect-related wave packets. Thus, the defect can be highlighted by accounting for the effect of noise and wave packet interference on the time-delayed signal array. However, when defects are located outside the sensor network, the limited information available may reduce the imaging performance. Numerical simulations and experimental studies conducted on plate-like structures demonstrate the proposed method achieves higher imaging clarity and localization accuracy for the internal defect compared with the external defect, with the former exhibiting mm-level absolute localization errors.