The problem considered is how to estimate the value of
the quantizer scale factor Q that corresponds to a given
number of bits B in a given frame of a sequence. The
solution exploits two statistical models B(Q) of I-, P-
and B-frames coded by an MPEG coder. These models
B(Q) depend on some parameters related to video
content. In the first model, there are very few such
parameters that can to be estimated from a previous
similar frame. In the second model, the content
parameters are the AC coefficients of DCT. Accurate
choice of the value of the quantizer scale factor Q can
be made using these two models. The experimental
results have been presented for MPEG-2 MP@ML
coders but some other experiments show that similar
approach is useful for other hybrid coders like H.263
and MPEG-4.