Den 25. feb. 2015 15:36, skrev Timothy B. Terriberry: > Harald Alvestrand wrote: >> I've come to the conclusion that varying resolution should be considered >> one tool for achieving lossy compression, and should not be a primary >> dimension - YMMV. > > Can you say a little bit more about how you would go about comparing the > results of two encoders that chose to encode a sequence (or even > individual frames within a given sequence) at different resolutions?
I think comparision tools that care about what the codecs do rather than what the result is are problematic - so I'm not sure whether this is different from all the other choices encoders can make. The overall process would have to return an image suitable for display - for simplicity of testing, I would make the output resolution the same as the input resolution. We've experimented with this sort of resolution change in the WebRTC pipeline - encoding in a lower resolution saves both CPU and bandwidth. (Just for fun, I integrated such a rescaling into the compare-codecs pipeline that encodes using the H.261 encoder, to allow comparing H.261 with its very limited available resolutions to other codecs on HD-sized clips - it "works", for some value of "work". psnr values of 35 dB where x264 achieves 40 dB - it seems psnr isn't particularly sensitive to the resulting blurriness). > > _______________________________________________ > 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
