[FFmpeg-devel] [PATCH 2/2] libavformat/mpegts: demux DVB VBI data

[Scott Theisen]

On 11/24/24 13:17, Marton Balint wrote:
>
>
> On Fri, 22 Nov 2024, Scott Theisen wrote:
>
>> On 11/22/24 16:12, Marton Balint wrote:
>>>
>>>
>>>  On Fri, 15 Nov 2024, Scott Theisen wrote:
>>>
>>>>  From: ulmus-scott <scott.the.elm at gmail.com>
>>>>
>>>>  DVB VBI data is defined in ETSI EN 301 775 and can include EBU 
>>>> teletext
>>>>  data
>>>>  as defined in ETSI EN 300 472.
>>>>
>>>>  ETSI EN 300 468 defines teletext_descriptor, VBI_data_descriptor, and
>>>>  VBI_teletext_descriptor, which has the same definition as, but 
>>>> different
>>>>  use
>>>>  from, teletext_descriptor.
>>>
>>>  Hmm, adding a new codec for VBI data might be OK, but I am not sure if
>>>  there is a better way to automagically let the framework know if 
>>> there are
>>>  teletext compatible VBI packets, so the teletext decoder can be used. 
>>
>> Would another AVCodecID AV_CODEC_ID_DVB_VBI_WITH_TELETEXT that would 
>> use the teletext decoder work?  Would this be AVMEDIA_TYPE_SUBTITLE 
>> even though it also includes data?  (Also, what teletext decoder?)
>
> I am not that big fan of adding both DVB_VBI and 
> DVB_VBI_WITH_TELETEXT, considering these use the same essence format.
>
> And VBI_teletext_descriptor is an additional descriptor to 
> VBI_data_descriptor as far as I understand the standard. (Although 
> ETSI EN 300 468 6.2.47 seems to be referring to teletext_descriptor as 
> an additional descriptor, but that makes little sense to me, they 
> meant VBI_teletext_descriptor here?)
>

I hadn't noticed that and that is very confusing.  I think it should say 
VBI_teletext_descriptor there.

>>
>>>  Or maybe we should simply set the codec id to DVB_TELETEXT if there 
>>> is a
>>>  VBI_TELETEXT_DESCRIPTOR, and make the teletext decoder parse all data
>>>  units in search for a comptaible data unit id?
>>>
>>
>> That would work only if you don't want to use any of the other VBI 
>> data, which is what MythTV currently does, ignoring packets where 
>> data_identifier is not EBU data and data_field()s where data_unit_id 
>> is not EBU teletext data as defined in ETSI EN 300 472.
>>
>> My main concern is if someone wants to decode the other types of VBI 
>> data how would they do that, or know to try, if it claims it is just 
>> EBU teletext data?
>
> FFmpeg does not have data decoders, so the user has to parse the 
> packets anyway if he is interested in VBI.
>
>>>  Because if we set the codec to VBI for VBI_TELETEXT cases, won't this
>>>  break the automatic probing of unknown codecs? (which otherwise might
>>>  recognize the stream as teletext)? So maybe for 
>>> VBI_TELETEXT_DESCRIPTOR
>>>  case we should keep the coded id NONE and rely entirely on probing?
>>
>> I have no idea how this would interact with the automatic probing.
>
> Me neither to be honest, and relying entirely on probing is not very 
> nice in the first place, so let's just forget this.
>
> Maybe we should stick to your original idea, use a single codec for 
> the VBI streams, but make it AVMEDIA_TYPE_SUBTITLE, considering it can 
> contain teletext (even if there is no VBI teletext descriptor in case 
> of inverted teletext) and even closed captions, which our decoders 
> might be able to decode to something meaninful.
>
> Then anybody interested can add a new decoder for this new codec ID 
> (which obviously can share code with libzvbi-teletextdec), so it will 
> be able to decode teletext as well.
>

Making the codec AVMEDIA_TYPE_SUBTITLE is fine with me.  Even though the 
VBI data is backwards compatible with the EBU teletext data, I think it 
should be a separate codec since AV_CODEC_ID_DVB_TELETEXT is known to 
contain *only* EBU teletext data.  (This information would be lost if 
they were combined into one codec.)

It is not clear to me if "Inverted Teletext" has any difference from EBU 
teletext data other than the inverted framing_code.

libzvbi-teletextdec should work unmodified with the new codec type since 
it already checks for the correct data_identifier and data_unit_id.

The automatic probing (libavformat/dvbtxt.c) would return 0 if any other 
data_identifier or data_unit_id values appeared, which I would assume 
would be true for any stream with a VBI_data_descriptor because 
otherwise the stream would only have EBU teletext data and should have a 
teletext_descriptor.

Regards,

Scott Theisen