At MPEG we use this to determine the codec performance for the "random access" case. I don't know why the 1 second interval was chosen and don't think that its relevant in many use cases, but MPEG codecs are aimed at satisfying a broad range of applications and so this test case has been maintained. Members of the potential IETF activity may find that this is an irrelevant test case.
Regards, On Tue, Mar 3, 2015 at 10:47 AM, Thomas Davies (thdavies) < [email protected]> wrote: > No, I suspect it was to be more appropriate for broadcast applications, > which usually have more frequent intra for channel switching. > > Back in the day, 1/2 second or 12 frame GOPs were deemed essential for > MPEG-2 digital TV to mimic analogue switching times, but I think that is > long gone - now audio is switched first and then the video when possible, > to give an illusion of immediacy. I think low bit rate channels could > easily have intra periods of 2 or 3 seconds. > > I agree it's short for streaming applications, and could be longer. > However, including a decent proportion of intra in the metrics is probably > good. > > Regards > > Thomas > > -----Original Message----- > From: Thomas Daede [mailto:[email protected]] > Sent: Monday, March 02, 2015 10:20 PM > To: Thomas Davies (thdavies) > Cc: [email protected] > Subject: Re: [video-codec] Test sequences, automated and otherwise > > Do you know if the 1-second keyframe interval was also for reasons of > compute time? (It allows each GOP to be encoded in parallel). It > unfortunately seems a bit short for many streaming applications, and also > prevents rate control from being tested. > > 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. > > On 03/02/2015 02:52 AM, Thomas Davies wrote: > > The "Class A" JCT-VC sequences were originated at 4k or 8k and cropped > > down for reasons of compute time. The 8k-originated ones (Nebuta and > > SteamLocomotive) are very noisy and were shot with an early camera and > > sensor. There wasn't much good 4k at the time HEVC started, and it > > would certainly be good to get more. > > > > regards > > > > Thomas > > > > On 25/02/15 17:44, Thomas Daede wrote: > >> The JCT-VC test set only includes three clips for video conferencing, > >> which I don't think is sufficient. It also specifies a intra frame > >> period of 1 second, which is not reasonable. It does not feature any > >> 2160p test sequences either, which is a bit strange. > >> _______________________________________________ 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 > -- Mohammed Raad, PhD. Partner RAADTECH CONSULTING P.O. Box 113 Warrawong NSW 2502 Australia Phone: +61 414451478 Email: [email protected]
_______________________________________________ video-codec mailing list [email protected] https://www.ietf.org/mailman/listinfo/video-codec
