[FFmpeg-devel] [PATCH] avfilter: Port mp=softpulldown to lavfi

[Paul B Mahol]

On 1/29/15, Michael Niedermayer <michaelni at gmx.at> wrote:
> On Thu, Jan 29, 2015 at 05:42:39PM +0000, Paul B Mahol wrote:
>> On 1/29/15, Michael Niedermayer <michaelni at gmx.at> wrote:
>> > On Thu, Jan 29, 2015 at 09:20:43AM +0000, Paul B Mahol wrote:
>> >> On 1/29/15, Michael Niedermayer <michaelni at gmx.at> wrote:
>> >> > On Wed, Jan 28, 2015 at 03:13:27PM +0000, Paul B Mahol wrote:
>> >> >> Signed-off-by: Paul B Mahol <onemda at gmail.com>
>> >> >> ---
>> >> >> Bit-exact with mp=softpulldown except first frame which is
>> >> >> uninitialized
>> >> >> data in mp filter.
>> >> >
>> >> >> TODO: pts needs update, workaround is to use setpts filter.
>> >> >
>> >> > patch LGTM except the pts / time_base issue
>> >>
>> >> mp=softpulldown sets all frames to MP_NOPTS_VALUE.
>> >>
>> >> I do not see way how to set filter to use variable frame rate, which
>> >> is possible as repeat_pict can change during stream.
>> >
>> > Heres my attempt to fix the issue
>> > The problem which i saw was that the first frames timestamp was
>> > stored in the temporary internal frame and then returned for many
>> > frames so they all had a incorrect and equal to the video starttime
>> > value
>> >
>> > Below this timestamp is reset to AV_NOPTS_VALUE and
>> > the code below takes the somewhat conservative approuch of only
>> > adjusting timestamps for the common 30000/1001 case and otherwise
>> > only passing untouched timestamps through or if they are off by a
>> > field period seting them to AV_NOPTS_VALUE if fps != 30000/1001
>> >
>> > timebases and frame rates are identical between input and output
>> > they needed no adjustment for the testcase i could find
>>
>> What testcase?
>
> interlaced_flag_switch.mpeg though this one contains some damaged
> frames so maybe its not the ideal testcase
> was just the first i found
>

But still:
>
>> The pattern can change to almost anything.
>

So I'm for removal of this filter.

> [...]
> --
> Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
>
> Old school: Use the lowest level language in which you can solve the
> problem
>             conveniently.
> New school: Use the highest level language in which the latest
> supercomputer
>             can solve the problem without the user falling asleep waiting.
>