Conventional Doppler techniques can only detect the axial component of blood flow. To obtain the transverse flow component, an approach based on the dependence of Doppler bandwidth on Doppler angle has been widely investigated. To compute the bandwidth, a full Doppler spectrum is often required. Therefore, this approach has not been applied to real-time, two-dimensional Doppler imaging because of the long data acquisition time. To overcome this problem, a correlation-based method is proposed. Specifically, variance of the Doppler spectrum is used to approximate the square of the Doppler bandwidth. Because variance is computed efficiently and routinely in correlation-based color Doppler imaging systems, implementation of this method is straightforward. In addition, the two-dimensional velocity vector can be calculated and mapped to different colors using the color mapping function of current systems. Simulations were performed, and experimental data were also collected using a string phantom with the Doppler angle varying from 23 degrees to 82 degrees . Results indicate that the correlation-based method may produce significant errors if only a limited number of flow samples are available. With averaging, however, the Doppler angles estimated by the correlation-based method can achieve good agreement with the true angles by using only four flow samples with proper variance averaging.