Digital Signal Processing Reference
In-Depth Information
horizontal and vertical frequencies, as human vision is less
sensitive to high frequency.
Further scaling is provided by a quantization scale factor,
which will be discussed further in the rate control description
section.
The DC coefficients are coded differentially, using the differ-
ence from the previous frame, which takes account of the high
degree of the average, or DC, level of adjacent blocks. Entropy
encoding similar to JPEG is used. In fact, if all the frames are
treated as I-frames, this is pretty much the equivalent of JPEG
compressing each frame of image independently.
14.5 Compressing P-frames
When processing P-frames, a decision needs to be made for
eachmacroblock, based upon themotion estimation results. If the
search for a match with another macroblock does not yield a good
match in the previous I- or P-frame, then it must be coded as I-
frames are coded
no temporal redundancy can be taken
advantage of. However, if a good match is found in the previous I-
or P-frame, then the current macroblock can be represented by
a motion vector to the matching location in the previous frame,
and by computing the residual, quantizing and encoding (inter-
coded). The residual uses a uniform quantizer, and this quantizer
treats the DC component the same as the rest of the AC
coefficients.
e
Residual Quantization Table
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
