Digital Signal Processing Reference
In-Depth Information
another frame. I.e., first the macroblock of the preceding frame may have
to be shifted to a suitable position with the aid of the motion vector of the
macroblock and then the difference with respect to the macroblock at this
position is calculated. Using the DCT, these 8 x 8 difference values are
then transformed into the frequency domain and then quantized. The same
also applies to the backward prediction of B pictures.
7.1.7 ZigZag Scanning with RunLength Coding of Zero
Sequences
After the zig-zag scanning (Fig.7.21.) of the quantized DCT coefficients, a
large number of adjacent zeroes is obtained. Instead of these many zeroes,
only their number is then simply transmitted by using run-length coding
(RLC) (Fig.7.22.), transmitting, e.g. the information 10 times 0 instead of
0, 0, 0, ...0. This type of redundancy reduction, in conjunction with DCT
and quantization, provides the main gain in the data compression.
1 7 3
6
0
0
- 1
0
2
0
- 2
0
0
0
0
0
0
- 1
0
0
0
0
0
0
0
0
0
0
0
0
0
- 1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
RLC
173, 6, 2*0, -1, 1*0, 2, 1*0, -2, 6*0, -1, 13*0, -1, 34*0
Fig. 7.23.
Run-length coding (RLC)
7.1.8 Huffman Coding
Codes occurring frequently in the RLC-coded data stream are also sub-
jected to Huffman coding (Fig.7.23.), i.e., the code words are suitably re-
coded, resulting in further redundancy reduction. In this type of coding, the
codes used frequently are recoded into particularly short codes as in Morse
code.
Search WWH ::
Custom Search