2016-11-16 17:18 GMT+01:00 Matteo Naccari <[email protected]>: >> (I am assuming here that libturing is both slower and produces worse >> quality than libx265, if this is not true, the following has of course little >> relevance.) > > According to the reference I've circulated before we achieve a lower > compression penalty but we are slower than libx265.
Not sure I understand. (Please correct me if the following is wrong / is not what you are saying:) You argue that to measure the quality of an encoder, it is not necessary to decode the output bitstream and compare with the decoded bitstream of other encoders but that it is sufficient to analyze and compare two bitstreams - what quantizers are used - and the filesizes (and the time encoding needed)? > Like I said this codec is a live project in continuous development. Which non-obsolete encoder project (including x264) is different? >> I don't think comparing to a reference implementation makes >> sense if better implementations exist and are used. > > The coding efficiency provided by the HEVC reference implementation > (i.e. HM) gives you the "theoretical" limit a standard can achieve. (Still not sure if I understand) Are you saying here that a relevant comparison between two implementations of a video standard is to let both encode a source file at the highest possible quality mode and then look which one achieved better quality? I am sorry that my knowledge about comparing video codecs is so limited: I thought you choose an input sample and an output filesize for which you assume a "good" encoding is possible but limitations will be visible (for the educated viewer). You then test if the implementations are able to reach the given filesize in an acceptable time (upper limit), then you decode both output files and compare them: Either (easy and uncertain) using psnr, ssim or something similar or (much more difficult but more exact) visually. That's how it was done over the last decade, no? [...] > http://www.bbc.co.uk > This e-mail (and any attachments) is confidential Please remove this from your mails! Carl Eugen _______________________________________________ ffmpeg-user mailing list [email protected] http://ffmpeg.org/mailman/listinfo/ffmpeg-user To unsubscribe, visit link above, or email [email protected] with subject "unsubscribe".
