[MEncoder-users] issues for time-lapse video encoding

[Hugh Secker-Walker]

Thanks for the quick feedback.

> >    which codec will degrade the images the least or have the simplest
> >    behavior/results?
> 
> Whatever one you're going to be using later.

So far, x264 gives the highest quality for a given final bitrate.

> Honestly though, I don't see why you don't fully compress the video on
> the fly.  With such low frame-rates, you should have ample CPU power.

Yes, CPU usage is almost nil during the initial capture.  My newbie
understanding is that the encoder can do a better job of allocating
bits to frames if it gets to see the stats from the first pass before
it does its heavy-duty compression work on the second pass.  So, I
don't see how to "fully compress" on the fly with a single pass.  Is
there some trick to this?

> >  BTW, I'm currently using the mjpeg codec, but there's no control that
> >  I can see (ffmpeg) for the compression that it's doing....
> 
> vqscale

Ah, that's from mencoder.  So far I've not been able to get mencoder
to tell the USB camera to run in 1280x1024, so I've been using ffmpeg
for the initial capture.

Nonetheless, I think you are suggesting to run the first pass with a
constant quality, conservatively high bitrate.  Then let the second
pass compress more agressively while being more adaptive.


> >    it would be nice to remove all the objects, e.g. birds, that appear
> >    in only a single frame. 
> 
> hqdn3d with a very, very high temporal value might work for you.  Or
> perhaps tinterlace + pp=lb + tfields.  Probably not good enough though.
> You probably need to find a more specialized tool for the job.

Thanks for the starting pointers.

-Hugh