AW: [XviD-devel] Transcode module

[peter ross]

>Hi,
>
> > Okay, sounds reasonable. So in theory more than one video_object 
>startcode
> > per frame is possible, even if not used in XviD (or anyother MPEG-4
> > codec I know of). It would therefore in general not be "clean" to simply
> > cut at VOP-start boundaries, but the timecode has to be taken into 
>account
> > to see which VOPs belong in one chunk. Right?
>
>This is not correct. A VO Startcode indicates the start of a new video
>object, obviously. Someone could insert in a video elementary stream (VES)
>another VO startcode indicating the start of a new video object (Could be
>useful for e.g. changing the dimension of the video in a streaming
>environment). But currently most decoders crash when seeing a video
>elementary stream with more than one VOS/VO/VOL header.

from my understanding of the iso specification, video-objects and
video-object-layers are stored sequentially in the bitstream.
vops can't be interleaved.

eg.    |vol1 vop1.1 vop1.2 vol1.3|vol2 vop2.1 vop2.2 vop2.3|

maybe if the vol is repeated, the vops can be interleaved.
this would be rather ineffecient.

eg.
       |vol1 vop1.1|vol2 vop2.1|vol1 vop1.2|vol2 vop2.2|vol1 \
             vop1.3|vol2 vop2.3|

>Anyhow, if you get hold on a Envivio coding station (costs approx. USD
>1500), you can combine multiple VESs into a MP4 file. Gives you fancy 
>things
>like video-in-video playback and much much more.

yep, but i wonder if it simply inserts the video into different
mp4 tracks, and use bifs to string it together.

-- pete


_________________________________________________________________
MSN Photos is the easiest way to share and print your photos: 
http://photos.msn.com/support/worldwide.aspx