Page 90 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience

Section 3.4. The H.263 Standard 67

layer that is used for upward prediction of an EI- or EP-picture may be an
I-picture, a P-picture, or the P part of a PB- or Improved PB-frame. Thus, an
EI-picture in an enhancement layer may have a P-picture as its lower-layer
reference picture, and an EP-picture may have an I-picture as its lower-layer
enhancement picture. For both EI- and EP-pictures, the prediction from the
lower reference layer uses no motion vectors. However, EP-pictures use mo-
tion vectors for the prediction from their prior reference picture in the same
layer.

c: Spatial scalability: Spatial scalability refers to enhancement information
used to increase the picture quality by increasing picture resolution either
horizontally, vertically, or both. Spatial scalability is very similar to SNR
scalability. The only di1erence is that before the picture in the reference layer
is used to predict the picture in the enhancement layer, it is interpolated by
a factor of 2 either horizontally or vertically (1-D spatial scalability) or both
horizontally and vertically (2-D spatial scalability). The interpolation /lters for
this operation are de/ned by the standard. Spatial scalability is illustrated in
Figure 3.6(c).
d: Multilayer scalability: It is possible not only for B-pictures to be tem-
porally inserted between pictures of types I, P, PB, and Improved PB, but
also between pictures of types EI and EP (whether these consist of SNR
or spatial-enhancement pictures). It is also possible to have more than one
SNR or spatial-enhancement layers in conjunction with a base layer. Thus a
multilayer scalable bitstream can be a combination of SNR layers, spatial lay-
ers, and B-pictures.

3.4.6.10Reference Picture Resampling Mode (Annex P)
In this mode, a resampling operation can be applied to the previously decoded
picture in order to generate a new warped picture for use as reference for
predicting the currently encoded picture. For example, if the previous reference
picture and the current picture are of di1erent source formats, then this mode
can be used to resample the previous picture to match the source format of
the current picture. Another example is to use this mode to warp the previous
reference picture to compensate for global motion. Warping and warping-based
motion estimation methods are discussed in Chapter 5.

3.4.6.11 Reduced-Resolution Update Mode (Annex Q)

This mode allows the encoder to send information encoded at a low resolution
to update a higher-resolution reference picture and produce a /nal picture
at the higher resolution. This mode is particularly useful when encoding a
highly active scene, and allows an encoder to increase the picture rate at