Page 102 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience

Section 3.5. The MPEG-4 Standard 79

transparent pel, c k = 1 for an opaque pel, N =10 pels for IntraCAE, and
N = 9 pels for InterCAE.

2. Determine the probability of the pel being transparent (or opaque) by
using the context number to index a table of probabilities de/ned by
the standard.
3. Use the indexed probability to drive an arithmetic encoder for codeword
assignment.
When all pels in the BAB have been encoded, the arithmetic encoder is
terminated.

3.5.2.2 Gray-Scale Shape Coding
A gray-scale alpha plane has a similar representation to the binary alpha plane,
with the di1erence that elements within the plane can take on a range of
values, usually 0 to 255 with 8-bit representation, designating the degree of
transparency of the corresponding pel. Gray-scale shape information consists
of two parts. The /rst part is the support information. This is obtained by
thresholding the gray-scale alpha plane at 0 (i.e., any value that is not equal
to 0 is set to 255). Support information is encoded using the binary shape
coding methods described previously. The second part of gray-scale shape
information contains the gray-scale values of the alpha plane. This is encoded
using methods similar to the texture encoding methods described later in this
chapter (Section 3.5.4).

3.5.2.3 Scalable Shape Coding
Besides changing the coding mode of BABs, additional mechanisms are em-
ployed for controlling the quality and bit rate of binary shape information. One
method is by reducing the resolution of the BAB by a factor of 2 or 4. The
resulting 8 × 8or4 × 4 BAB is encoded using any of the available modes. At
the decoder, the reduced-resolution BAB is /rst decoded and then upsampled.
Another method for reducing the binary shape bit rate is by changing the ori-
entation of the BAB. The e,ciency of the CAE algorithm can depend on the
orientation of the BAB. In some cases, transposing the BAB before coding it
can increase coding e,ciency. In this case, the decoder decodes the BAB and
then transposes it back to its original orientation.

3.5.3 Motion Estimation and Compensation

Motion estimation and compensation methods in MPEG-4 are very similar
to those employed by other standards. The main di1erence is that block-