Den 03. mars 2015 03:19, skrev Mo Zanaty (mzanaty): > Harald will probably disagree, but the convention is to disable rate > control for comparing objective metrics.
Oh no, I completely agree that this is the convention. I just don't agree that it's useful :-) Rate control is normally an > application-specific configuration of the encoder which can add noise or > bias if comparing different rate controls. Rate control strategies can > often be abstracted to apply to any modern codec (with adaptive > quantizers), so it is best to remove this source of noise/bias if you > can¹t guarantee identical (not default) strategies across the comparison. > Note that some GOP structures often include some basic rate controls by > default, like different QP offsets for hierarchical B/P frames. This can > also be a source of noise/bias unless configured identically across the > comparison. > > Mo > > On 3/2/15, 5:19 PM, Thomas Daede <[email protected]> wrote: > At the moment, I run all codecs with rate control on in their constant > quality mode. Perhaps this is not the right thing to do. > > _______________________________________________ > video-codec mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/video-codec > _______________________________________________ video-codec mailing list [email protected] https://www.ietf.org/mailman/listinfo/video-codec
