Loading...

mythtv-users@mythtv.org

[Prev] Thread [Next]  |  [Prev] Date [Next]

Re: [mythtv-users] Keyframe identification in cutlist editor John Pilkington Wed Feb 22 10:00:22 2012

On 22/02/12 15:23, Nick Morrott wrote:
> On 22 February 2012 11:12, Mike Perkins<[EMAIL PROTECTED]>  wrote:
>
>> Having recently done some manual editing using avidemux, I can tell you that 
>> if
>> your cuts are /not/ on keyframes the end result will likely be noticeable on
>> screen as random blocks, hashes or other constructs.
>>
>> Makes sense really, only keyframes are complete frames, which means that when
>> one partial frame (the one you cut on) is followed by another partial frame 
>> (the
>> next one you cut on, but following a different keyframe) the difference
>> information is too great to be resolved. Only when the /next/ keyframe comes
>> along can the GPU/TV resolve everything correctly.
>
> I use avidemux (after a pass through ProjectX) quite a bit and always
> ensure the start of a new sequence starts on an I- (or key-) frame (as
> any MPEG-2 stream Group of Pictures (GOP) should). Sequences end on P
> or I frames, but these usually crop up in the right place anyway,
> especially if the video fades to black over a second or two. I've not
> seen any bad blocking artefacts as a result of not having /both/ the
> start and end of a sequence fall on an I-frame - the starting I-frame
> is the key one.
>
> Some video editors (I don't think avidemux is amongst these) will
> reconstruct affected GOPs after lossless cutting to ensure the
> sequence of I,B and P frames at cutpoints allow successful
> reconstruction of the video without blocking/distortion (which occurs
> when a sequence of frame cannot be successfully rendered).
>
> Cheers,
> Nick
>

It's not clear to me exactly how or why you're using ProjectX followed 
by avidemux; I suppose the aim is to get more sophisticated effects. 
But I ought to point out that when PX is used by my script or in 
MythArchive to make the cuts it also makes the joins.  I'm afraid I 
haven't investigated how it does that - and of course it's described by 
its developers as educational software to allow the user to understand 
the basics of MPEG video processing - but I haven't seen artefacts at 
the joins; just clean jump cuts at positions that aren't quite frame 
accurate.

Of course mythtranscode offers the --allkeys option, intended to allow 
clean editing.  I haven't used that so I have no idea how robust it is, 
but it sounds as if the output is a fully decompressed version of the 
original file.

Cheers,

John


_______________________________________________
mythtv-users mailing list
[EMAIL PROTECTED]
http://www.mythtv.org/mailman/listinfo/mythtv-users