Speckle tracking is capable of searching the complete vector velocity but is susceptible to the frame-to-frame image decorrelation. With plane-wave excitation (PWE), the high frame-rate reduces the scatterer displacement and thus increases the signal correlation. Nevertheless, the lower image quality with PWE may suffer from spatial velocity gradient outside the sample volume and reduce the achievable accuracy of speckle tracking. In this study, we propose a recursive PWE compounding technique to improve the image quality for accuracy of vector velocity estimation while the high-frame-rate nature of PWE for vector velocity estimation is retained. In the proposed compounding, several PWE images from different steering angles are combined coherently to suppress the beam sidelobes at the cost of longer acquisition time. The coherent PWE compounding is further combined with a recursive technique to generate high-resolution images with different features at a frame rate of pulse repetition frequency; thus retaining the high frame-rate nature. The neighboring high-resolution image pairs with the same image feature are utilized for velocity detection with speckle tracking. Extensive simulations were performed to test the recursive PWE compounding method for its performance in speckle-tracking based vector velocity estimation. Results indicate that, the lateral estimation is more susceptible to the spatial flow velocity gradient than the axial estimation is. When the Doppler angle is 0°, it is shown that the mean axial velocity remains close to the ideal parabolic profile while the detection of lateral velocity is markedly interfered by the flow gradient. In the case of large Doppler angle, it is also evident that an accurate lateral estimation is achieved with the coherent PWE compounding method. Note that the standard deviation with the PWE compounding is generally smaller as compared to the conventional PWE image, indicating the improved robustness in both latera- - l and axial estimation. Nevertheless, the error of lateral estimation is still larger than that of axial estimation. It is demonstrated that the recursive PWE compounding with speckle tracking is capable of improving estimation of vector velocity with high frame rate. In addition to vector velocity estimation, the proposed method can also be potentially applied to elastography.