[Libav-user] Criteria for h.264 key-frame detection

[Robert Krüger]

On Mon, Oct 21, 2013 at 5:46 PM, Carl Eugen Hoyos <cehoyos at ag.or.at> wrote:
> Robert Krüger <krueger at ...> writes:
>
>> Since only P- and I-slices are parsed (I can see that
>> in the debug output), I would indeed expect every
>> I-frame (meaning any frame only consisting of one
>> I-slice in the stream) to be a valid access point
>
> So you mean a P-frame after an I-frame is not allowed to
> reference an earlier I-frame?
> I don't think this is correct.

You are probably correct here. I have to check the spec. Would be a
very weird stream, though. Have you seen such a beast in the wild?

Anyway, my theory is that it is not the case here and, since this is
an AVCHD file which is usually accompanied by other auxiliary files, I
guess that explicit marking of random access points in the stream is
usually not necessary for hard/software that relies on the AVCHD spec
being followed, because they are stored in a separate file (my
assumption, for which I have no proof). The bad thing is that I
haven't found a version of the spec online and it costs a lot of
money, so I cannot check this immediately. Anyway, it does not help
much in deciding how ffmpeg is supposed to decide it is a random
access point.

Btw, is it intentional that you don't answer questions I ask in my
emails? It feels a bit impolite to be honest. I don't remember not
answering a question you have asked.