Den 24. sep. 2015 21:12, skrev Thomas Daede: > On 09/24/2015 01:07 AM, Harald Alvestrand wrote: >> Uh...... perhaps we should bite the bullet and just set rate control >> parameters? > > I'm not sure what you mean here - you mean specify quantizers as the bounds?
I've argued for a long time that a codec that can't be rate-controlled is useless for interactive Internet usage - and that therefore, it makes sense to actually use rate controls when we test the codec. We're now faced with a situation where we have specified test conditions that specify certain bitrate targets - but the only means we have of hitting those bitrate targets is to encode the file over and over again until we find something that has a bitrate that matches our target. If we could, instead, specify the bitrate target when encoding the file, and measure the resulting quality metrics on the file after bitrate controls are applied, we would have, in my opinion, a test scenario that more resembles real-life usage. Of course, fixed-quantizer encodings are reasonable to use when trying to measure the impact of codec tweaks that don't interact with the rate controls. But that's not the only thing we want to measure. > > _______________________________________________ > 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
