Page 86 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience

Section 3.4. The H.263 Standard 63

should be resized to /t into a smaller part of the displayed video picture,
which should then be kept unchanged.

3. Picture Snapshot: The mode can be used to signal that the current
picture, or a speci/ed rectangular part of it, is labeled for external use
as a still-image snapshot of the video content.

4. Video Time Segment: The mode can be used to signal the beginning and
the end of a speci/ed subsequence of video data to be used externally.

5. Progressive Re4nement Segment: The mode can be used to signal the
beginning and the end of a speci/ed subsequence of video data. Rather
than being a continually moving scene, this subsequence of video in-
cludes a start picture followed by a sequence of zero or more pictures
to re/ne its quality.
6. Chroma-Keying Information: The mode can be used to indicate that
the chroma-keying technique is used to represent transparent and semi-
transparent pels in the decoded video pictures. When being presented
on the display, transparent pels are not displayed. Instead, a background
picture is revealed that is either a prior reference picture or an exter-
nally controlled picture. Semitransparent pels are displayed by blending
the pel value in the current picture with the corresponding value in the
background picture.

3.4.6.7 Improved PB-Frames Mode (Annex M)

This mode represents an improvement compared to the original PB-frames
optional mode (annex G). The main di1erence between the two modes is that
the original PB-frames mode can utilize only bidirectional prediction to predict
the B part in a PB-frame, whereas the improved PB-frames mode can utilize
forward, backward, or bidirectional prediction.
The bidirectional prediction method is the same as in the original PB-frames
mode, except that in this case no delta vector is transmitted. In the forward-
prediction method, a B macroblock is predicted from the previously decoded
P-picture and a forward motion vector is transmitted. In the backward-
prediction method, a B-macroblock is predicted from the corresponding
P-macroblock currently decoded in the same PB-frame, and therefore no back-
ward motion vector needs to be transmitted.
This mode signi/cantly improves coding e,ciency in situations in which
downscaled P-vectors (utilized in the original PB-frames mode) are not good
candidates for B-prediction. In particular, the backward prediction is useful
when there is a scene cut between the previous P-frame and the current PB-