I found that both Bezier curves and B-splines are described with a formula
$p(t)=\sum\limits_{i=0}^d B^i_m p_i$
but in the case of B-splines $B^i_m$ are B-spline blending functions, while for Bezier curve these are Bernstein polynomials.
Also I found, that perspective transform can easily be applied to NURBS, which is similar to B-spline but each control point is associated with weight, which can also be represented in homogeneous coordinates
$p_i = \left( \begin{array}{c} x_iw_i \\ y_iw_i \\ w_i \end{array} \right)$
And the plot will be the follows:
$p(t)=\frac{\sum\limits_{i=0}^d B^i_m p_i}{\sum\limits_{i=0}^d B^i_m w_i}$
(is this correct?)
My question is: how important is replacing B-spline blending function to Bernstein polinomials?
Can I transform Bezier curves by applying the same technique?
And also direct formulas would be appreciated :)
UPDATE
Also I am not sure about summation indice. In Bezier formula, the order of polynomial coincides with the number of control points, while in B-spline is not.
Even on Wolfram's page the formula for "rational Bezier curve" is given with free order index:
Does this mean that Stephen also not sure about polynomials? :)