Abstract Ultrasonic color flow mapping (CFM) is employed for the diagnosis of a variety of cardiovascular abnormalities. But CFM permits only the analysis of the in-beam component of the velocity distribution. A reconstruction method based on a geometrical approach was developed to calculate the two-dimensional, in-plane velocity vector using the information of two independent CFM measurements obtained from two closely spaced transducers. However, with the CFM measurements being limited by the noise and quantified in a limited number of velocity values, the accuracy of the velocity vector reconstruction is not satisfactory. A simple filtering process was applied to improve the results. The method was tested with simulations and a series of in vitro measurements. The results of the simulations were compared with analytical solutions. The in vitro measurements were compared with the results of laser Doppler anemometry (LDA) and theoretical calculations. The results of in vivo measurements were compared with theoretical calculations based on Womersley's theory. The comparison showed good applicability of the method to different flow fields. The flow vector reconstructed without filtering had an error of more than 20%; whereas our method, including the filtering process, had an error of less than 10%, in comparison with the theoretical or the LDA results. The method could be extended to three-dimensional flows.