[FFmpeg-user] Problem with vf pad ... help please

[Tim Nicholson]

On 17/09/12 16:36, Nicolas George wrote:
> Le jour de la Vertu, an CCXX, Mark Himsley a écrit :
>> You are falling into the trap that the all of the pixels on a 720x576
>> line are part of the active 4:3 picture. Where as only 702 pixels are.
>>
>> If you wish to ignore the calculated scale and pad values I gave you
>> above, and insist on sticking with your calculated numbers, then you
>> (like many before you and many to follow) will get the will get the
>> images 2.5% too narrow. I'm sure that will flatter most people, being
>> 2.5% too thin, but it is wrong. Please believe the experience of 26
>> years at the BBC.
> 
> Do you have any normative reference about that? It seems to contradict what
> I remember (I'll have to dig into the standards again) reading.
> 

You have to work it out from ITU-R BT.601-7 TABLE 4 and ITU-R BT.470-6 !

This gives (for 625/50):-

Number of samples per total line for each signal =864.

Analogue total line timing 64us
Analogue active line timing 52us

Therefore digital samples per active analogue line = 864*52/64 =>702.

The 4:3 aspect ratio dates back to analogue days and so aspect ratio
calculations must be based on the analogue picture width and height, i.e
702x576.

Hope this helps.

> Note that I do not doubt your 26 years of BBC, just I wonder who is doing
> something wrong: it could also be the ones producing the padded 702 pixels.
> 
> Regards,
> 
> [..]


-- 
Tim