Cryptography Reference
In-Depth Information
compensation, adaptive quantization, enhanced reference picture selection,
inverse DCT mismatch reduction, de-blocking and de-ringing filters, as well
as error concealment. Its Annex X also defines Profiles and Levels to group
optional modes and parameters to improve inter-operability above the baseline
capability.
2.5.6 MPEG-4
The MPEG-4 standard is fundamentally different in nature from its predeces-
sors. It represents the scene as the composition of video objects, rather than
just TV-like raster scan digital signal. In MPEG-4, the smallest entity in a
picture is not a pixel, but an object with its associated shape, texture, and
motion. Thus MPEG-4 is often referred as the second generation video coding
standard, or the object based video coding standard.
The audio, video, graphics and other multimedia components of MPEG-4
are known as objects. These can exist independently, or a few of them can be
grouped together. The result of grouping objects is an MPEG-4 scene. The
strength of this object-oriented approach is that any of the components can
be easily manipulated, optimally represented and interacted. Fig. 2.18 shows
an example of how an MPEG-4 video scene can be segmented into a number
of video objects and the background.
The advantages of the object based coding is its similarity to the HVS,
easier to interact with objects in the picture and good potential for use in
various applications. However it is very di cult to properly define, segment
and track the video objects, and needs very complicated interface to conven-
tional display. Fig. 2.19 shows how an MPEG-4 object can be decoded and
composed to reconstruct the original video scene, and Fig. 2.20 shows the
MPEG-4 scene graph structure of Fig. 2.19.
Besides the object based coding, MPEG-4 has provided a number of ad-
vanced tools for shape coding, motion estimation and compensation, texture
coding, error resilience, sprite coding and scalability, such as,
• Shape coding: can be performed in binary mode or greyscale mode.
• Motion coding: 1616 or 88 block based, variable object boundaries,
overlapped block motion compensation, half-pixel resolution.
• Texture coding: based on 88 DCT, variable object boundaries, coe cient
prediction.
• Wavelet transform based still texture coding.
• Sprite coding, which is an extended, panoramic image used for background
pictures. This extended image is coded and transmitted in advance. On
the receiver side, an image patch is extracted from the extended image
and is used as a background picture of the decoded image.
• Model based coding.
• Face object coding.
• Mesh object coding.
Search WWH ::




Custom Search