The MPEG-4 Part 2 Codec - A Practical Guide to Video and Audio Compression - page 228

Table 12-1 New Terminology Introduced With MPEG-4

Term Meaning

VS Visual sequence. The top level of the hierarchical object-based representation.

VO Visual object. These might be video sequences or static images. A VS object

owns them.

VOL Video object layer. These represent a layer within the display that contains

some video. A VOL is owned by a VO.

GOV Group of VOPs. These are similar in concept to the GOP idea but are extra

layers of structure. VOL objects own GOV objects.

VOP Video object plane. A single video sequence playing back in its own plane of

existence. A GOV object is the parent object when the structure is visualized as
a hierarchical tree.

manufacturers talk about them as if they are the Holy Grail, but wavelets are just another
technique for encoding macroblocks.
MPEG-4 uses wavelet encoding with the discrete wavelet transform (DWT), but only
for static textured images. It still uses DCT for compression of natural-motion video.
The difference between the Fourier-based DCT algorithm and the wavelet algorithm
is minor but quite important. In DCT transforms, the higher frequencies are sampled
across the same space so there are more cycles of the higher-order components. Wavelet
encoding works on the principle of having a fixed number of cycles regardless of the fre-
quency being used. So as the frequency increases, the area being considered gets smaller.

12.4 Further Improvements Introduced in MPEG-4 Part 2

This has yet to make a major impact in terms of deployments. Some manufacturers of
mobile pocket devices are implementing video capabilities with MPEG-4 part 2. As a
result, some very neat, compact, and cheap digital video cameras that store video on mem-
ory cards rather than tapes are starting to appear.

12.5 Coding Efficiency Improvements

When the MPEG-4 standard was revised to version 2, some new tools were introduced,
which have a profound effect on coding efficiency. These tools include the following:

Global motion compensation (GMC)

●

Quarter-pixel motion compensation (QPMC)

●

Shape-adaptive discrete cosine transforms (SADTC)

●