Two-dimensional blood velocity estimation has shown potential to solve the angle-dependency of conventional ultrasound flow imaging. Clutter filtering, however, remains a major challenge for large beam-to-flow angles, leading to signal drop-outs and corrupted velocity estimates. This work presents and evaluates a compounding speckle tracking (ST) algorithm to obtain robust angle-independent 2-D blood velocity estimates for all beam-to-flow angles. A dual-angle plane wave imaging setup with full parallel receive beamforming is utilized to achieve high-frame-rate speckle tracking estimates from two scan angles, which may be compounded to obtain velocity estimates of increased robustness. The acquisition also allows direct comparison with vector Doppler (VD) imaging. Absolute velocity bias and root-mean-square (RMS) error of the compounding ST estimations were investigated using simulations of a rotating flow phantom with low velocities ranging from 0 to 20 cm/s. In a challenging region where the estimates were influenced by clutter filtering, the bias and RMS error for the compounding ST estimates were 11% and 2 cm/s, a significant reduction compared with conventional single-angle ST (22% and 4 cm/s) and VD (36% and 6 cm/s). The method was also tested in vivo for vascular and neonatal cardiac imaging. In a carotid artery bifurcation, the obtained blood velocity estimates showed that the compounded ST method was less influenced by clutter filtering than conventional ST and VD methods. In the cardiac case, it was observed that ST velocity estimation is more affected by low signal-to-noise (SNR) than VD. However, with sufficient SNR the in vivo results indicated that a more robust angle-independent blood velocity estimator is obtained using compounded speckle tracking compared with conventional ST and VD methods.