This work concerns the development of a methodology the objective of which is to characterize and diagnose two-phase flow regime transitions. The approach is based on the fundamental assumption that a transition flow is less stationary than a flow with an established regime. During the first time, the efforts focused on: (1) the design and construction of an experimental loop, allowing to reproduce the main horizontal two-phase flow patterns, in a stable and controlled way; (2) the design and construction of an electrical impedance probe, providing an imaged information of the spatial phase distribution in the pipe; and (3) the systematic study of the joint time–frequency and time-scale analysis methods, which permitted to define an adequate parameter quantifying the unstationarity degree. During the second time, in order to verify the fundamental assumption, a series of experiments were conducted, the objective of which was to demonstrate the correlation between unstationarity and regime transition. The unstationarity degree was quantified by calculating the Gabor’s transform time–frequency covariance of the impedance probe signals. Furthermore, the phenomenology of each transition was characterized by the joint moments and entropy. The results clearly show that the regime transitions are correlated with local time–frequency covariance peaks, which demonstrates that these regime transitions are characterized by a loss of stationarity. Consequently, the time–frequency covariance constitutes an objective two-phase flow regime transition indicator.