MPEG-2 Video Coding - Digital Video and Audio Broadcasting Technology - page 110

discern fine structures in a picture, e.g. thin lines, as well as coarse struc-
tures. This is precisely the main point of attack for data reduction methods
like JPEG and MPEG. However, irrelevance reduction is always associ-
ated with an irretrievable loss of information which is why the only
method considered in data processing is redundancy reduction as, e.g. in
the well known ZIP files.

Data reduction

Redundancy reduction
(no loss of information)
reversible

Irrelevance reduction
(loss of information)
nonreversible

Fig. 7.1. Data reduction

In MPEG, the following steps are carried out in order to achieve a data
reduction factor of up to 130:

• 8 bits resolution instead of 10 bits (irrelevance reduction)

• Omitting the horizontal and vertical blanking interval (redun-

dancy reduction)

• Reducing the color resolution also in the vertical direction

(4:2:0) (irrelevance reduction)

• Differential pulse code modulation (DPCM) of moving pictures

(redundancy reduction)

• Discrete cosine transform (DCT) followed by quantization (ir-

relevance reduction)

• Zig-zag scanning with variable-length coding (redundancy re-

duction)

• Huffman coding (redundancy reduction)

Let us begin again with the analog video signal from a television cam-
era. The red, green and blue (RGB) output signals are matrixed to become
Y, C B and C R signals. After that, the bandwidth of these signals is limited
and they are analog/digital converted. According to ITU-BT.R601, this
provides a data signal with a data rate of 270 Mbit/s. The color resolution
is reduced in comparison with the brightness resolution, making the num-
ber of brightness samples twice that of the C B and C R values and resulting