View Single Post
Old October 12th 18, 06:57 PM posted to alt.windows7.general
external usenet poster
Posts: 1,078
Default question for the video editor folks

In article ,

"Paul" wrote in message
Editing a video on GOP boundaries for MPEG2 only, isn't
the limit of the science. But for "frame accurate" editing
without re-rendering, the software still has to mess up
the contents of at least one GOP (up to 12 frames,
one would hope only 5 frames would get messed up).
Messed up means "recomputed and corrected" to mesh
with the GOP on either side. I've not seen any
critical reviews of how well this works. Whether
a discerning customer would choose to re-render
the whole thing, as a "better for quality" solution.

I can confirm that VideoRedo manages frame-accurate edits, with no
discernable loss of picture quality (eg macroblock pixellation or JPEG
fringing) around the edit point, irrespective of whether the edit points are
on GOP boundaries. The only exception that I have found to this is for
low-bitrate and low-resolution channels (544x576 rather than 720x576, at 1
Mbps) when *by default* there can be a few frames with corruption - but
there is a fix which imposes a minimum bitrate (I use 2 Mbps) for any
regenerated frames.

To the best of my knowledge, the first frame of the second clip is generated
as a full frame by calculation from the last full frame before the edit
point plus any difference frames, and then new difference frames are
generated from that point onwards to the next full frame in the source.

I would say that as long as the new frames are free of any blemishes, this
is a much better solution than re-rendering the whole project, both because
it is faster and because you don't lose any quality in recoding any more
frames than you have to.

I would expect all similar software to adopt this approach; I was rather
gobsmacked to see how bad AVIdemux was at joining two videos when the join
isn't on a GOP boundary - very bad macroblock pixellation.

I'm unsure why no one has mentioned using VirtualDub. I often strip the
start/ends of video files and that's pretty straightforward doing a
direct stream copy. Mind you it'd be a bit of a pain to get all the junk
stripped out of say an hour long tv show but it'd do it and I believe
that it can be done without re-encoding. Putting the peices back
together is trivial.