A dynamic complex electromagnetic induction imaging technique is developed with the aid of the linear Kalman filter (LKF) for real-time reconstruction of molten metal flow. The forward problem is solved by an edge finite element method. The inverse problem is treated as a state estimation problem. The conductivity distribution is estimated with the aid of the Kalman estimator. The Kalman gain matrix is pre-computed and stored off-line to minimize the on-line computational time. Simulation and experimental tests are reported to illustrate the reconstruction performances in the sense of spatio-temporal resolution in a simplified geometry of the metal flow in continuous casting.