Am 04.02.2015 um 12:31 schrieb Timothy B. Terriberry:> M. Pabis wrote: >> 1. Each thread deals with frames from intra frame up to next intra frame >> - 1; > > This works if you know where the intra frames are.Could this information be gathered by having one thread encode a downsampled version of the input video sequence, or would this be a bad predictor? Who knows, perhaps one can also gather data on relative bitrate distribution between segments. Maik
Maik Merten wrote:> Am 04.02.2015 um 12:31 schrieb Timothy B. Terriberry: > Could this information be gathered by having one thread encode a > downsampled version of the input video sequence, or would this be a bad > predictor?Yes. x264 has a lookahead thread with exactly this purpose (among others). But current libtheora has nothing like this.> Who knows, perhaps one can also gather data on relative bitrate > distribution between segments.That requires significantly more lookahead, of course.
On Wed, Feb 4, 2015 at 1:16 PM, Timothy B. Terriberry <tterribe at xiph.org> wrote:> Maik Merten wrote: > > Am 04.02.2015 um 12:31 schrieb Timothy B. Terriberry: > > Could this information be gathered by having one thread encode a > > downsampled version of the input video sequence, or would this be a bad > > predictor? > > Yes. x264 has a lookahead thread with exactly this purpose (among > others). But current libtheora has nothing like this. > > Great, thanks for comments and ideas.Is there anything obvious I'm missing why this improvement would be hard or a waste of time? Mateusz Pabis -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.xiph.org/pipermail/theora-dev/attachments/20150205/c02dd1a5/attachment.htm