Page 78 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience

Section 3.4. The H.263 Standard 55

3.4.5 Optional Coding Modes
There are four optional coding modes that can be signaled at the picture level.
These modes are de/ned in annexes to the standard and are brieJy described
next.

3.4.5.1 Unrestricted Motion Vector Mode (Annex D)
In this mode, motion vectors are allowed to point outside the reference picture
area. When a pel pointed to by a motion vector is outside the reference picture
area, an edge pel is used instead. This edge pel is found by limiting the
motion vector to the last full-pel position inside the reference picture area.
Limitation of the motion vector is performed on a pel basis and separately for
each component of the motion vector. In this mode also, the range for motion
vector components is extended to [−31:5; +31:5], with the restriction that if the
predictor is in the range [−15:5; +16], then only values that are within a range
of [−16; +15:5] around the predictor can be reached. If, however, the predictor
is outside [−15:5; +16]; then all values within the range [−31:5; +31:5] with
the same sign as the predictor can be reached. Allowing motion vectors to point
outside the reference picture area improves prediction along picture edges in
the case of camera or background movement. This is particularly useful for
small picture formats (where border MBs represent a high percentage of the
picture area). The extended motion vector range allows better prediction for
large picture formats and a high amount of movement.

3.4.5.2 Syntax-Based Arithmetic Coding Mode (Annex E)
In this mode, all VLC Hu1man coding=decoding operations of H.263 are
replaced with arithmetic coding=decoding operations. As already discussed in
Section 2.5.5, arithmetic coding removes the restriction of representing each
symbol by an integral number of bits, achieving more coding e,ciency but at
the expense of more computational complexity.

3.4.5.3 Advanced Prediction Mode (Annex F)
This optional mode includes two advanced prediction techniques: the use of
four motion vectors per MB, and the use of overlapped motion compensation
(OMC). In addition, this mode allows motion vectors to point outside the
reference picture area. If this mode is used in combination with the unrestricted
motion vector mode, then the motion vectors will also have an extended range.
If the mode is used in combination with the PB-frames mode, then OMC is
only used for P-pictures, not for B-pictures.
In this mode, the encoder makes a decision (which is not de/ned by the
standard) whether to transmit one motion vector or four motion vectors per