[FFmpeg-devel] FFv1.3: full GOP 1/300, 2-pass statistics for Derf's collection

Peter B. pb at das-werkstatt.com
Sun Dec 9 23:33:58 CET 2012


On 12/05/2012 06:12 AM, Loren Merritt wrote:
> On Wed, 5 Dec 2012, Peter B. wrote:
>
>> It's very interesting to see that 2-pass with GOP=1 reaches almost the
>> same compression ratio as single-pass with GOP=300. In some cases, it's
>> even small than GOP=300 (marked red).
> Of course. Each of your test videos is a single scene. The probability
> tables don't need to change much from one frame to the next, because the
> content doesn't change much.
>
> The same result *might* also hold for a whole movie, but your experiment
> doesn't predict that one way or the other.
I've now ran it on material from our archive. Some of it is available
online [1].
They're from VHS and Digibeta, and definitely contain several scenecuts,
etc.

The results match those of Derf's collection and can be seen here:
http://download.das-werkstatt.com/pb/mthk/ffv1_stats/latest/ffv1_gopdiff-mthk-yuv.html
http://download.das-werkstatt.com/pb/mthk/ffv1_stats/latest/ffv1_sizediff-mthk-yuv.html


Regards,
Pb

== References:
[1] http://download.das-werkstatt.com/pb/mthk/FFv1-Examples/


More information about the ffmpeg-devel mailing list