Page 80 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience

Section 3.4. The H.263 Standard 57

Figure 3.4: Prediction in PB-frames mode

For prediction of a B-block, both forward, MV F , and backward, MV B ,
motion vectors are needed. Those are not transmitted but are derived at the
decoder by scaling the corresponding P-block motion vector, MV, using the
temporal resolutions of the P- and B-pictures with respect to the previous
P-picture. The derived motion vectors can be optionally enhanced using a
transmitted delta vector MVDB. The forward (or backward) motion vectors
for chroma blocks are derived by summing the corresponding luma forward
(or backward) motion vectors, dividing by 8, and then rounding to the nearest
half-pel resolution. To be able to predict a B-macroblock, the correspond-
ing P-macroblock is /rst reconstructed. For pels of a B-block where MV B
points outside the reconstructed P-macroblock, forward prediction using MV F
and the previous decoded P-picture is used. However, for pels of the same
B-block where MV B points inside the reconstructed P-macroblock, bidirec-
tional prediction is used. In this case, the prediction is the average (with
truncation) of the forward prediction, using MV F and the previous decoded
P-picture, and the backward prediction, using MV B and the reconstructed
P-macroblock. This process is illustrated in Figure 3.4. With this mode, the
frame rate can be increased without a signi/cant increase in bit rate.

3.4.6 H.263, Version 2 (H.263+)

Version 2 of the H.263 standard is informally known as H.263+. This version
adds a number of optional feature enhancements to version 1. In the process
of adding these new features, the precise de/nition and requirements of the
original version 1 syntax and semantics were not changed. In fact, version 2
is backward compatible with version 1. The additional optional feature set can
be summarized in terms of the new types of pictures, a modi/ed unrestricted