This saved my evening ;)

Initially I've got points from OpenFOAM sample in the plane, then I collected all Xes and Ys, and then (cause there was also fluid with channel involved and velocities was not available at every point and when data is not available then it's zero)...

>> vels = importdata('vels.dat');
>> I = ~(vels(:, 3) == 0 & vels(:, 4) == 0);
>> quiver(vels(I, 1), vels(I, 2), vels(I, 3), vels(I, 4))

As all these points for zero values gone, figures look much clearer.