[FFmpeg-user] Decimate question

Mark Filipak (ffmpeg) markfilipak at bog.us
Wed Dec 30 20:02:37 EET 2020

On 12/30/2020 12:46 PM, Paul B Mahol wrote:
> On Wed, Dec 30, 2020 at 6:44 PM Mark Filipak (ffmpeg) <markfilipak at bog.us>
> wrote:
>> On 12/30/2020 03:35 AM, Paul B Mahol wrote:
>>> On Wed, Dec 30, 2020 at 6:08 AM Mark Filipak (ffmpeg) <
>> markfilipak at bog.us>
>>> wrote:
>>>> Decimate will drop a repeated frame in each sequence of 5 (=default)
>>>> frames. I assume the dropped
>>>> frame can be any 1 of the 5. Is there a way to discover which of the 5
>>>> frames it dropped?
>>> -v debug
>> May I make a feature request? Could it be added as ... is it called side
>> data?
>> I haven't used side data yet so I'm not quite sure what it is and how it's
>> used.
>> The frame # would need to be readable by follow-on logic inside
>> expressions.
>> This is for processing videos that have previously been telecined, a frame
>> dropped, and a frame
>> repeated, in various sequences. The sequence of the video destruction is
>> not important. What's
>> important is the phase of the decimation within the modulo-5 in the video
>> that remains. In 4 of the
>> 5 cases (phases), it's actually possible to almost completely recover the
>> original video (prior to
>> the wrecking). 4 out of 5 cases involve the loss of just 1 field in 10.
>> The 5th case involves the
>> total loss of a whole frame and is not recoverable (but since that case
>> would have an amazing amount
>> of judder, it's unlikely to be found in the wild). However, I have found
>> other cases of this video
>> species in the wild.
> Sorry, but you make nonsense entries, very often and as expected.

I have the evidence. I have made short captures of the f'ked up videos and how to fix them.
Before decimation, there's 5 cases (one of which is unrecoverable). After decimation, the 4 
remaining cases can't be differentiated. If I knew which frame# decimation had removed, then I could 
differentiate between them and apply the best recovery of each of the 4 video species.

More information about the ffmpeg-user mailing list