[x264-devel] Behaviour of Annex B encoding: bug or not?

[Jason Garrett-Glaser]

2010/1/7 Måns Rullgård <mans at mansr.com>:
> Måns Rullgård <mans at mansr.com> writes:
>
>> Jason Garrett-Glaser <darkshikari at gmail.com> writes:
>>
>>>> It's only because x264 generates more complete session information, which
>>>> includes several near-zero multi-byte integer parameters that result in
>>>> escape-required patterns of traffic, that we are seeing the problem.
>>>> It doesn't occur when the units talk to each other.
>>>
>>> This is unrelated.  Simply due to random chance, statistically,
>>> escaping bytes will be needed about once every 4 million bytes of
>>> data--ignoring headers.
>>
>> Billion, surely?
>
> Or not.  But 16 million, not 4.

Escape bytes are required for the sequences:

0x000000
0x000001
0x000002
0x000003

These sequences occur 4 / 2^24 of the time, or  1 / 2^22 of the time,
or roughly 1 / 4 million of the time.

Dark Shikari