Databases Reference
In-Depth Information
where
P
1
through
P
6
correspond to the six different blocks in the macroblock, and is one if
the corresponding block has a nonzero quantized coefficient and zero otherwise.
19.5.5 Rate Control
The binary codewords generated by the transform coder form the input to a transmission buffer.
The function of the transmission buffer is to keep the output rate of the encoder fixed. If the
buffer starts filling up faster than the transmission rate, it sends a message back to the transform
coder to reduce the output from the quantization. If the buffer is in danger of becoming emptied
because the transform coder is providing bits at a rate lower than the transmission rate, the
transmission buffer can request a higher rate from the transform coder. This operation is called
rate control
.
The change in rate can be accomplished in two different ways. First, the quantizer being
used will affect the rate. If a quantizer with a large step size is used, a larger number of
coefficients will be quantized to zero. Also, there is a higher probability that those not quantized
to zero will be one of the values that have a shorter variable-length codeword. Therefore, if
a higher rate is required, the transform coder selects a quantizer with a smaller step size, and
if a lower rate is required, the transform coder selects a quantizer with a larger step size. The
quantizer step size is set at the beginning of each GOB, but can be changed at the beginning of
any macroblock. If the rate cannot be lowered enough and there is a danger of buffer overflow,
the more drastic option of dropping frames from transmission is used.
The ITU-TH.261 algorithmwas primarily designed for videophone and videoconferencing
applications. Therefore, the algorithm had to operate with minimal coding delay (less than
150 milliseconds). Furthermore, for videophone applications, the algorithm had to operate at
very low bit rates. In fact, the title for the recommendation is “Video Codec for Audiovisual
Services at
p
s,” where
p
takes on values from 1 to 30. A
p
value of 2 corresponds to
a total transmission rate of 128kbps, which is the same as two voice-band telephone channels.
These are very low rates for video, and the ITU-T H.261 recommendations perform relatively
well at these rates.
×
64 kbit
/
19.6 Model-Based Coding
In speech coding, a major decrease in rate is realized when we go from coding waveforms to
an analysis/synthesis approach. An attempt at doing the same for video coding is described
in the next section. A technique that has not yet reached maturity but shows great promise
for use in videophone applications is an analysis/synthesis technique. The analysis/synthesis
approach requires that the transmitter and receiver agree on a model for the information to be
transmitted. The transmitter then analyzes the information to be transmitted and extracts the
model parameters, which are transmitted to the receiver. The receiver uses these parameters to
synthesize the source information. While this approach has been successfully used for speech
compression for a long time (see Chapter 18), the same has not been true for images. In a
delightful topic,
Signals, Systems, and Noise—The Nature and Process of Communications
,
published in 1961, J.R. Pierce [
11
] described his “dream” of an analysis/synthesis scheme for
