h264 saturation point
- DJ_Izumi
- Joined: Wed Oct 03, 2001 8:29 am
- Location: Canada
- Contact:
h264 saturation point
Just curious if someone can suggest to me at what bitrate h264 would become saturated at and additional bandwidth would do no good towards improving quality. (For the purpose of HD-DVDs and Blu-Ray disks)
- Willen
- Now in Hi-Def!
- Joined: Sun Jul 10, 2005 1:50 am
- Status: Melancholy
- Location: SOS-Dan HQ
I'm sure it will depend a lot on the material being encoded and the resolutions and framerates used. At the moment, since there are so little HD-DVD™ and Blu-ray Disc™ authoring tools aside from ridiculously expensive commercial boxes and the occasional cut-down consumer hardware/software package, it is hard to judge what is acceptable. Especially since there is exactly 1 standalone player for each format available in North America for purchase today, testing your home recorded Blu-ray Disc is not a cheap or easy task. This is assuming you have the money for the Pioneer BDR-101A internal burner, or either of Sony's VAIOs - the VGN-AR190G or VGC-RC310G. Not to mention blank Blu-ray Disc media. And forget about your own recorded HD-DVDs right now, the PC recorder that NEC announced isn't shipping yet.
- Zarxrax
- Joined: Sun Apr 01, 2001 6:37 pm
- Contact:
First of all, just to make things clear, "saturated" typically means the codec maxes its quality out and is not able to make the filesize as large as you wanted it to be. With xvid, this can happen quite often, especially if you have the minimum quantizers limited to 2.
H.264 is unlikely to become "saturated" per se, such as xvid, because it allows for much higher video quality, and will gladyly suck down more and more bitrate, to the point of eventually making the video lossless.
You are probably asking more along the lines of "what is the highest quality where you wont see any more noticable gain?" Well like willen said, it will vary from source to source. It has been said that h.264's quantizer 18 is roughly equivalent to xvid's quantizer 2. Most people accept that as being the best that you need, but you can get visual artifacts at that compression level, so lower quantizers may be called for in some cases.
H.264 is unlikely to become "saturated" per se, such as xvid, because it allows for much higher video quality, and will gladyly suck down more and more bitrate, to the point of eventually making the video lossless.
You are probably asking more along the lines of "what is the highest quality where you wont see any more noticable gain?" Well like willen said, it will vary from source to source. It has been said that h.264's quantizer 18 is roughly equivalent to xvid's quantizer 2. Most people accept that as being the best that you need, but you can get visual artifacts at that compression level, so lower quantizers may be called for in some cases.
- Zero1
- Joined: Fri Jan 02, 2004 12:51 pm
- Location: Sheffield, United Kingdom
- Contact:
Re: h264 saturation point
You cannot use bitrate as a guide of quality. It's all about quantizers. You could have a really complex source that would look shit at 800kbps, but on the other hand a not so complex source may look fine at 800kbps. The complex source would have to use a higher average quantizer for the stream to keep to the same average bitrate. A higher quantizer means more course rounding of the DCT co-efficients, which causes more distortion (such as blocking).DJ_Izumi wrote:Just curious if someone can suggest to me at what bitrate h264 would become saturated at and additional bandwidth would do no good towards improving quality. (For the purpose of HD-DVDs and Blu-Ray disks)
For example I encoded .hack roots at 86MB and it looked decent, but the average quantizer was around 20, which is pretty good, so it obtains it's quality at a low filesize. Other sources like Gundam X require more bits because it is harder to motion compensate. The same quantizer for Gundam X might be around 160MB. Remember, it's the quantizer that basically controls the quality. Stuff like motion estimation and whatnot help to predict what happens in other frames, and this means less energy in the residual, which in turn means compressing it at a set quantizer would end up using less bits than if you used crappy motion estimation and was left with a ton of residual.
A lot of people term Q2 in XviD saturation, well it isn't, but its the most sane trade between quality and filesize. If you want a guide, you can use CRF or QP 18 in x264. QP 0 is lossless, so you can't really be specific about a "saturation point".
I've just seen Zarx's post, there isn't much more to add.
7-zip // x264 (Sharktooth's builds) // XviD (Koepi's builds) // MP4box (celtic_druid's builds) // Firefox // CCCP