[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[OpenDivX] OpenDivX Encoder -> OpenDivX Daily Forums Digest

OpenDivX Encoder -> OpenDivX Daily Forums Digest

Topic:		Does anybody want this?
Author:		gruel
Posted:		 2001-02-23 14:23
On 2001-02-22 19:24, czw wrote:
I have a gigahertz Athlon myself, but I am connected to a network with
at least 200 students. There are several computers in the network that
don't do much (Linux file servers), and that would gladly do some DivX
encoding in their spare time. On the other hand, quite a few of the
computers around here are K6:s, Durons and Athlons.

Actually both projects do not interfere with each other at all. All
optimization (AMD/Intel/RISC) will be part of the encode-library. To
create a distributed version would only make sense outside the encoder
itself, say in the application that calls it.
I guess it could be less than a days work to turn an existing mpeg2avi
or similar into a version using MPI or PVM (I like MPI more), running on
a cluster of Linux/NT-PC or Workstations, if documentation is available.
After that a more advanced step would be possible: Two-pass-encoding,
load balancing depending on processor speed and scene complexity etc. 

The more I think about it, the more I like the idea: Since with a
shared-lib mechanism it would always use the most recent encode-lib
present, it could be a great tool for optimization. Not for speed,
because running a cluster is surely the exception for users, but for
_quality_ improvements. 
Since you get almost a linear speedup when just splitting a video into n
parts for n processors, you could run several tests of different
parameters over night (when most machines are not used anyway). 

Hey, I like that idea! 
Now who has a documented encoder by hand to turn into a cluster-version?

How about the MPlayer? Is A'rpi reading this? 


Topic:		Auto fast/slow motion detection in encoder?
Author:		stobo
Posted:		 2001-02-23 14:38
On 2001-01-25 10:29, Anonymous wrote:
Microsoft uses a new method in its new WM8 codec to produce better image
quality in high motion scenes: 2-pass . The encoding process is done
twice. In the first run the media is analyzed for high and low motion
scenes. That is to set the best highest possible bitrate for each scene.
The low motion scenes don't require such a high bitrate so more space is
available for the high motion scenes. I've tested it, and I think it is
a really good method for long films that should fit on 1 CD.

So, to come to a conclusion: what about intergrating such a feature in

i thought the parameter on how many frames to averate the bitrate on
does just this.. during this number, the quantization is changed to
slowly react to the differences in encoding efficiency. quantization is
the absolute quality of one frame, so when there is a slow scene, the
codec ends up using less bits per frame 'cause motion prediction already
gets all the quality..

try to set this number to something like 10000 (or more for a full

am i wrong? (should go look at virtualdub's bits/frame graph to verify
this.. maybe some player also can show bitrate variations over time.)[
This message was edited by: stobo on 2001-02-23 14:42 ]

Topic:		Main differences between MPEG2 and DivX/MPEG4 ?
Author:		stobo
Posted:		 2001-02-23 14:47
On 2001-02-12 15:36, u08194 wrote:
>From what I saw, most of the difference is due to H673 quantizer being
used instead of the MPEG quantizer.  

MPEG4 is also supposed to support adaptive quantization matrixes and
quarter pell motion prediction.  None of these are currently
implimented.  May be adaptive quantizer will be implimented as part of

i thought that the longer keyframe interval plus wider motion prediction
ranges also have something to do with this..

and decoder accuracy! the encoder can really assume that minor
modifications are still valid if the frame data has been updated 100
times since the last key block. then again accuracy might be part of the
quantizer spec.

Topic:		Awesome quality!! but porblems with overlay driver in
Author:		stobo
Posted:		 2001-02-23 14:50
On 2001-02-12 10:15, musicvid wrote:
Btw i cant see any codec other than the mp3 codec in properties on
mediaplayer but dont think thats the problem since it plays openDivX

i believe the story is simple as that there is not a directshow filter,
just the old-fashioned avi codec, and windows media player is stupid
enough not to use the overlay driver in this case.. removing the filter
from any divx or other decompressor causes just the same.

bottom line: get a better player software 

Topic:		OpenDivX Variable Bitrate Encoding
Author:		stobo
Posted:		 2001-02-23 15:05
On 2001-01-25 08:05, Sandman wrote:
all around they write about "true Variable Bit Rate" ..what about a
"false Constant Bit Rate"? I mean: you set the overall medium bitrate of
the movie/clip/video and the encoder uses it freely to achieve constant
quality in all scenes..you should say to the encoder "I have a clip 45
min long and I want it to fit EXACTLY a 650Mb CD, doesn't matter if a
frame uses 1 byte and another one 1 Mb" ..obviously to do so the encoder
should have a "preprocessing step" that takes almost the same time of
the real compression, in order to decide how many bytes assign to each
frame...Divx4 has some flexibility in this sense, but it was tought on
the range of few frames, I think...
Anyone has ideas about such a thing?
(I'm not a programmer,unluckly)


i thought the min/max quantization and the average bitrate over n frames
parameters were just about this.. try setting the frame count very high
and see what happens

OpenDivX mailing list
[email protected]

Reply To Poster

Local References / HOW-TO / FAQs