| |
 |
|
|
Science Forum Index » Compression Forum » Video part of MPEG2 and MPEG4
Page 1 of 1
|
| Author |
Message |
| Nash |
 Posted: Wed Sep 17, 2003 8:50 pm |
|
|
|
Guest
|
Dear All,
Who can tell me why MPEG4 has higher compression rates than MPEG2?
What features of MPEG4 lead to this?
Best Regards,
Nash |
|
|
| Back to top |
|
| Alan Burk |
| Posted: Thu Sep 18, 2003 8:45 am |
|
|
|
Guest
|
There are two MPEG-4 algorithms. Compression rate of MPEG-4 Part 2 is not
much higher than MPEG-2. MPEG-4 Part 10 (also called Advanced Video Coding)
has lots more coding tools such as intra prediction modes etc., different
block sizes, Content Adaptive Binary Arithmetic Coding, etc. etc. It isn't
any one particular coding tool that makes the difference, it's the
combination of all of them.
"Nash" <r88072@csie.ntu.edu.tw> wrote in message
news:eda9c155.0309171850.3100db3e@posting.google.com...
Quote: Dear All,
Who can tell me why MPEG4 has higher compression rates than MPEG2?
What features of MPEG4 lead to this?
Best Regards,
Nash |
|
|
| Back to top |
|
| nicobi |
| Posted: Fri Sep 19, 2003 9:23 am |
|
|
|
Guest
|
Quote: Who can tell me why MPEG4 has higher compression rates than MPEG2?
Because it's more complex ! Because it has taken into account more
than ten years of research and that the possibility that offer a
computer today is much better than the one used to develop the mpeg2
standard !
Quote: What features of MPEG4 lead to this?
well has said Alan :
- cabac is better than usual entropy lossless compression tools
(arround 10 to 20% !!) (like UVLC or Huffman like used in mpeg2) .
- Instead of having the choice between one or two macroblock types
(16x16 and 8x8). You have the choice between 7 types or so : 16x16
16x8 8x16 8x8 8x4 4x8 4x4 : so more precision.
- Hadamard add a precision regarding to standard sad metrics : this
implies a better correlation between coefficient, and a more efficent
dct/quantif step.
- more than one reference frame, so that for each motion estimation
of each block (regardless to his type), the algorithm can choose
betwenn the frames. This could be pretty good if you are using a
preanalyse tool that choses for you the most likely reference frames
(with some color similarity measure or else).
Well hope this helps.
nicobi |
|
|
| Back to top |
|
| |
|
Page 1 of 1
All times are GMT - 5 Hours
The time now is Sat Nov 22, 2008 7:50 pm
|
|