|
Compression Techniques used by Gigawave:
MPEG 2
MPEG 2 was primarily designed to provide the
most efficient transport for the terrestrial distribution of digital
television to the home. It incorporates highly sophisticated motion
compensation,
and as such is the most efficient of the coding systems in terms of quality-to-bit
rate ratio.
Benefits of MPEG 2:
 |
Highly efficient in terms of quality-to-bit rate ratio |
|
Scalable from 1.5 - 80Mbits with a many 'Level' (source
format) and 'Profile' (coding
process or complexity) options |
Gigawave MPEG 2 Options:
Of the possible options (six profiles and four levels) available Gigawave
have selected those permutations of 'Level' and 'Profile' which
fulfill the principal requirements of broadcasters, as follows:
Extremely Robust RF
MPEG at 6Mbits enables the use of QPSK, a particularly
robust form of modulation which, together
with the benefit of ½ rate FEC,
can operate within a single COFDM ensemble, and therefore
a standard 8
MHz RF channel. The result is high quality robust RF, which is ideal
for radio cameras and mobile links in difficult and/or harsh
environments.
Best Quality
MPEG in the range 18-24 Mbits gives top end quality. Using 64 QAM at ½ rate
FEC it can operate within a single COFDM ensemble, and therefore
a standard 8 MHz RF channel. This is ideal for point to point MVL links
where link budgets are under control, and where the inherent delay of
2-7 frames
(80-280 ms) can be tolerated.
How Does MPEG work?
Firstly, frame based spatial compression is
applied using Discrete Cosine Transform (DCT). DCT operates on blocks
of the picture, usually 8x8
pixels, resolving them into frequencies and amplitudes. MPEG can then
access this information and represent it in a far more efficient form,
for instance, not having to keep on repeating similar information,
and by removing all the zeros. The process produces what is known as
the Intra-frames or I-frames. I-frames contain
all the information to construct a whole picture.
Secondly, temporal compression is
applied, simply by looking for repeated information from frame to frame,
and also looking for where blocks
of information have moved ie motion estimation and compensation.
MPEG looks
at a number of frames, typically 12, and compares them. This group
of frames is referred to as the Group of
Pictures (GOP) Assessment
is made
by producing inter-frames, and there are two kinds of interframes frames:
Predictive frames P-frames, and Bi-directional predictive frames B-
frames
P- frames
P-Frames contain less data than I-frames,
and are generated by looking at the difference between the present frame
and the previous one.
B-frames
B-Frames
contain less data than I-frames, and are generated by looking
at the difference between the previous frame and the next frames
in the picture sequence.
Group of Pictures (GOP)
I-frames, B-frames, and P-frames are arranged in a sequence
to produce a GOP. A typical 12 frame sequence would be as follows:
A GOP always commences with a refresher I-Frame,
except for transmission where it is played out ahead of the last
two B-frames:
Technical information on MPEG Encoding
Wavelets
Gigawave has invested three years of R & D
on a line based compression system using Wavelets, operating at 18 Mbit/s
3/4 FEC.
Benefits of Wavelets Encoding:
|
Good quality |
|
Near instantaneous, system delay less than 1ms |
|
Gradually degrades with increased bit error rate (BER) |
|
Near instantaneous recovery from loss of signal |
Wavelets is of particular interest to Gigawave as
it is the only near instantaneous compression system. It is therefore
invaluable for radio cameras, and other transmission sources where integration
with other cameras and facilities makes any delay intolerable. [Compression
for radio cameras never has to cope with complete picture changes or cuts]
How does Wavelets work?
Wavelets compression passes the signal through a series of Wavelet filters
from which an assessment is made on a line by line basis. Wavelet filters
are orthogonally based which allows for perfect reconstruction.
The Wavelet
transform employs banks of Wavelet filters configured to resolve
the signal
into 'spatial frequency selective sub-bands' and produces a series
of coefficients. The coefficients follow a pattern (lots of zeros)
reflecting the redundancy from within the original signal.
Finally, the information is processed (compressed) using the following
techniques before it is transmitted:
|
Some filtering
of raw video |
|
Adaptive filtering |
|
All sync information removed |
|
Run length encoding - to remove runs of zeros |
|
Packing |
|
Adaptive quantisation |
|
1D or 2D processing |
A Guide to Digital Microwave Technology:
Digital Compression Techniques
Digital Modulation Techniques
Digital RF Transmission
A Glossary of Terms |
