[FFmpeg-user] ntsc-dvd doesn't default to interlaced.

[Andy Furniss]

Paul B Mahol wrote:
> On 7/27/13, Andy Civil <andycivil at gmail.com> wrote:
>> *never mind*
>>
>> Sorry, I think I found the answer by trial and error.
>>
>> c:\ffmpeg\bin\ffmpeg -i E:\temp\2013-06-06-hdp.mp4 -vf
>> scale=720:480,tinterlace=interleave_bottom -flags ildct+ilme -target
>> ntsc-dvd
>> -acodec mp2 -ac 2 -b:a 128000 2013-06-06-i.mpg
>>
>> This LOOKS right in VLC. It seems that if I put the filter definition first,
>> and
>> scale before interleaving, I get a good output.
>
> Yes, scaling interlaced content is no no.

I agree that weave scaled is one of the most annoying things you may see 
"out there" but -

ffmpeg seems to be able to scale interlaced as long as you tell it -

-vf scale=720:576:interl=1 -vcodec mpeg2video -flags +ildct+ilme -target 
pal-dvd  -aspect 16:9 -b:v 8000k

looks like it works on a 1080i h264 sample I just tested.

One wierd thing is that I got that from a post ages ago and it had an 
extra argument -

-vf scale=720:576:interl=1,fieldorder=tff

now whether I use tff or bff it has the effect of trashing the chroma, 
which may not be instantly obvious depending on content.

Of course my sample was 420 - I don't know what the OPs is, so that 
would be something to consider if he wanted to scale direct instead of 
de then re interlacing.

709 vs 601 colour is also something to consider as is the fact that if 
you scale down video you should ideally low pass filter it first - maybe 
more so for interlaced verticaly, I guess the anamophism sort of handles 
the kell factor horizontally.

I haven't got a clue how to do either with ffmpeg - TBH my test still 
looks OK anyway. I guess deinterlacing (on playback) helps soften and my 
sample is UK DVB-T broadcast which generally seems to be softer than 
kell factor would dictate to start with.




_______________________________________________
ffmpeg-user mailing list
ffmpeg-user at ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-user