The time of flight (TOF) features of Lamb waves have been widely applied to locate gross damage in plate structures, which may greatly facilitate the localization process when a sparse sensor network is adopted. Through a time-frequency analysis, this study extends the usage of TOF to nonlinear Lamb waves, taking advantage of their higher sensitivity to small-scale fatigue damage. The precision and practicality of using temporal signal features of linear and nonlinear Lamb waves for fatigue damage characterization are compared. Case studies are conducted on aluminum plates through numerical simulation and experiments, where miniature piezoelectric wafers are networked into a sparse array for Lamb wave excitation and acquisition. A probability-based imaging algorithm is then proposed to visualize damage. Finally, the comparison has motivated a synthesized scheme for locating fatigue damage in various conditions, enhancing the adaptivity of structural health monitoring for real-world applications.