Databases Reference
In-Depth Information
Motion
texture
coding
+
DCT
Q
Video
multiplex
-
Q
-1
Inverse
DCT
+
Predictor 1
Frame
store
Predictor 2
Predictor 3
Motion
estimation
Shape
coding
F I GU R E 19 . 18
A block diagram for video coding.
video coding algorithm can also use a background “sprite”—generally a large panoramic still
image that forms the background for the video sequence. The sprite is transmitted once, and
the moving foreground video objects are placed in front of different portions of the sprite based
on the information about the background provided by the encoder.
The MPEG-4 standard also envisions the use of model-based coding, where a triangu-
lar mesh representing the moving object is transmitted followed by texture information for
covering the mesh. Information about movement of the mesh nodes can then be transmitted
to animate the video object. The texture coding technique suggested by the standard is the
embedded zerotree wavelet (EZW) algorithm. In particular, the standard envisions the use of
a facial animation object to render an animated face. The shape, texture, and expressions of
the face are controlled using facial definition parameters (FDPs) and facial action parameters
(FAPs). BIFS provides features to support custom models and specialized interpretation of
FAPs .
The MPEG-2 standard allows for SNR and spatial scalability. The MPEG-4 standard also
allows for object scalability, in which certain objects may not be sent in order to reduce the
bandwidth requirement.
19.13 Packet Video
The increasing popularity of communication over networks has led to increased interest in the
development of compression schemes for use over networks. In this section we look at some
of the issues involved in developing video compression schemes for use over networks.
