[FFmpeg-devel] Cooperative coding between (N)LE and FFmpeg; possibly quality improvements?

[Stefan de Konink]

On Sat, 7 May 2011, Kieran Kunhya wrote:

>> I think you miss understood the 'scenecut' here. When in a
>> NLE clips are organised and placed on a timeline, it is
>> pretty known at what time which clip is played. To x264 this
>> information is only available from an image - not on meta
>> data level, which I would like to use to hint the encoder.
>
> The encoder will calculate scenecut costs based on the best compression
> efficiency (i.e which modes use the fewest bits). This trumps any metadata
> from the NLE. If you still think you can do a better job than the encoder
> you can force frametypes or change the scenecut thresholds.

This is a general approach which may (very well) work better. If your 
claim is that a black box method, outperforms any meta data approach in 
performance and in quality, it maybe better not to purse the idea.


Stefan