H264 1080p Bitrate... just how high does bitrates need to be!?

[liquidskin76]

Hi,

So, i've backed up Avatar to H264 MKV. I've create 3 backups using CQ22, CQ20, CQ18.

When checking the bitrate during heavy scenes, the bitrate is about...

CQ22... 12000mbps
CQ20... 15000mbps
CQ18... 20000mbps

I've sat up close to my Sony 40" lcd and can't see any difference between CQ22 and CQ18! Bear in mind that my lcd is only 720p so i can't judge at 1080p.

Now i know beauty is in the eye of the beholder and all, and i should just be going with CQ22 (as i can't see any difference), however i'm one of those guys who get bugged by knowing there's a physical difference!

For instance, there's a massive file size difference. Plus the extra bitrate. So there must be a difference!! Am i not seeing any difference due to viewing at 720p? I plan on one day getting a 1080p set so would i see a difference then?

I know h264 is one amazing codec at low bitrates. So my question is just how high does H264 1080p bitrate need to be? Is the 20000 upper range overkill? Is 12000 lower range sufficient?

My gut feeling is that 6000-8000 is fine as that seem to do for people when backing up to DVD9. BD bitrates i think are serious overkill!

Would love to hear peoples thoughts, to put my mind at rest!

Thanks

[orion44]

I tried to backup Avatar from the original DVD9 to 1400 MB DivX video.
That didn't work out too well.

So I just watch the DVD9 instead.

For example, that 20-minute fight scene in the air, when the blue people with their flying
horses are fighting the marines, requires DVD-like bitrates (over 4000 kbit/s) to look good on DivX.

So I guess it's the same case with Blu-ray backups.

[liquidskin76]

I tried to backup Avatar from the original DVD9 to 1400 MB DivX video.
That didn't work out too well.

So I just watch the DVD9 instead.

For example, that 20-minute fight scene in the air, when the blue people with their flying
horses are fighting the marines, requires DVD-like bitrates (over 4000 kbit/s) to look good on DivX.

So I guess it's the same case with Blu-ray backups.

What resolution is that at? DVD, 720p, or 1080p?

[7ekno]

I've sat up close to my Sony 40" lcd and can't see any difference between CQ22 and CQ18! Bear in mind that my lcd is only 720p so i can't judge at 1080p.

Best way to spot differences is interleave source + encode and step through it frame by frame, or use a tool like AvsPmod and bring up both files and flip back and forth between the tabs looking for differences ...

Keep in mind the human visual system is unique to each of us ... some of us can only pick differences at 16 fields per second, some of us can find those differences at 30 fields per second ... some of us can only distinguish 12 shades of the same color, while others can see over 24 shades .... point being, if you can't see the difference between CRF22 and CRF18, but are worried their might be one on a 1080p display, split the difference and go with a CRF20 encode ;)

I know h264 is one amazing codec at low bitrates. So my question is just how high does H264 1080p bitrate need to be? Is the 20000 upper range overkill? Is 12000 lower range sufficient?

My gut feeling is that 6000-8000 is fine as that seem to do for people when backing up to DVD9. BD bitrates i think are serious overkill!

Not all H.264 encoders are the same ;) Don't make the mistake of assuming the encoders used at the Studio are as efficient as x264 :P

The best bitrate is the one that provides a visual quality you are happy with :P If that's 6000kbps or 12000kbps, so be it :P

As for Xvid and the DVD9, first it would be wise to encode it anamorphic (so you aren't losing a heap of vertical resolution), and secondly, Xvid requires 63% more bitrate than x264 to get to the same "quality" (reference is the 2010 MSU Codec Comparison, page 21) ...

7ek

[orion44]

What resolution is that at? DVD, 720p, or 1080p?

DVD resolution, with cropped black borders.
(720x3xx, something like that if I remember correctly).

[nm]

For instance, there's a massive file size difference. Plus the extra bitrate. So there must be a difference!! Am i not seeing any difference due to viewing at 720p? I plan on one day getting a 1080p set so would i see a difference then?

That's easy to test by cropping the video on playback so that you get 1:1 pixel mapping. Just crop an area of 1280x720 pixels from the center of the frame with ffdshow or MPlayer, for example.

If you want to see what the differences are exactly, go with 7ekno's suggestion.

[Superb]

Avatar can be encoded really nicely to DVD9 size (~8GB) when using 720p (encoded w/ x264) w/ DTS core (extracted w/ eac3to).
Video bit-rate would be about 5525kb/s on the region A/1 blu-ray.

[orion44]

As for Xvid and the DVD9, first it would be wise to encode it anamorphic (so you aren't losing a heap of vertical resolution), and secondly, Xvid requires 63% more bitrate than x264 to get to the same "quality" (reference is the 2010 MSU Codec Comparison, page 21) ...

7ek

Thank you for this info.

[orion44]

Can someone tell me how big (or how small) is the difference in picture quality
between the original Blu-ray and it's backup encoded with x264?

In most cases, how does the backup compare? Nice, good, excellent, or awesome?

Where can I find picture comparisons between the Blu-ray and the x264 encoded backup?
Could someone post a couple of comparisons?

[Superb]

It depends if you re-sized the video (lowering the resolution to 720p, for example), what bit-rate you're using (which affects final encoding size), what x264 encoding settings you're using (which affects encoding time), and how you measure "difference".

[mp3dom]

Hi,
I've sat up close to my Sony 40" lcd and can't see any difference between CQ22 and CQ18! Bear in mind that my lcd is only 720p so i can't judge at 1080p.

Are you searching for differences during playback or differences on a frame-per-frame basis? If you want to see differences, watch both videos (source/compressed) on a calibrated monitor. You'll see them especially on 'delicate' areas (grain retention, color smoothing, dark areas etc)


Am i not seeing any difference due to viewing at 720p? I plan on one day getting a 1080p set so would i see a difference then?

Human eyes are not able to see fine differences or small macroblocks when there are 24fps. Unless moderate to high distortion occour (i.e. huge macroblocks or small macroblocks for more than 3-4 frames) or other severe artefacts (color banding, posterization, etc)


I know h264 is one amazing codec at low bitrates. So my question is just how high does H264 1080p bitrate need to be? Is the 20000 upper range overkill? Is 12000 lower range sufficient?

You should use as much as you can. In a frame-by-frame comparison even 40 Mbps isn't overkill.


BD bitrates i think are serious overkill!

Could be overkill if you're searching for strong artefacts during playback. I don't think is overkill if you want to be as close as possible to the original master.

Don't make the mistake of assuming the encoders used at the Studio are as efficient as x264 :P
The studio encoders are not as efficient as x264 at those middle-low bitrates, but at high bitrates they're more close to the original master (not all studio encoders, but at least the top two).

[7ekno]

Can someone tell me how big (or how small) is the difference in picture quality
between the original Blu-ray and it's backup encoded with x264?

In most cases, how does the backup compare? Nice, good, excellent, or awesome?

Depends on the settings you use ;)

Most would consider CRF18 - CRF20 "transparent" on 1080p content, meaning "no humanly detectable loss in quality" ... most of the time these settings result in a final filesize of 8-12GB (so one half to one third the original file size without a detectable loss of quality) ...


Where can I find picture comparisons between the Blu-ray and the x264 encoded backup?
Could someone post a couple of comparisons?

Not much point, transparent is transparent :)

Personally, I hate the artificial digital grain added by studios and the sometimes blurry appearance ... so these examples aren't aimed at "transparency", they are aimed at pleasing my eyes (load both pics in separate tabs, switch back and forth between tabs to spot differences ... pay attention to the guys hair which is sharper in the processed one and the removal of the grainy appearance below the airconditioner in the background to the right) ...

Original (http://www.users.on.net/~tekno/AVC/original1080p.png) - Processed (http://www.users.on.net/~tekno/AVC/processed1080p.png)

These are a fraction of the same 1080p frame, the encode was done via x264 --crf 19 --preset veryslow --tune film ...

7ek

[orion44]

Depends on the settings you use ;)

Personally, I hate the artificial digital grain added by studios and the sometimes blurry appearance ... so these examples aren't aimed at "transparency", they are aimed at pleasing my eyes (load both pics in separate tabs, switch back and forth between tabs to spot differences ... pay attention to the guys hair which is sharper in the processed one and the removal of the grainy appearance below the airconditioner in the background to the right) ...

Original (http://www.users.on.net/~tekno/AVC/original1080p.png) - Processed (http://www.users.on.net/~tekno/AVC/processed1080p.png)

These are a fraction of the same 1080p frame, the encode was done via x264 --crf 19 --preset veryslow --tune film ...

7ek

Wow, the backup looks awesome.
The ear, the hair and the forehead looks sharper.

By the way, what software players do you recommend
to watch x264 encoded content that can also play DTS audio?

[Blue_MiSfit]

The studio encoders are not as efficient as x264 at those middle-low bitrates, but at high bitrates they're more close to the original master (not all studio encoders, but at least the top two).


That's total crap. Provide an example of where x264 is worse than any other encoder at any bitrate, not just "high" bitrates.

@ OP:
I wouldn't stress about it too much. "Enough" is when you're happy.. :)

Personally I wouldn't suggest encoding most sources to 720p unless they don't actually contain detail at 1080p. Avatar is a source that contains LOTS of detail at 1080p - i.e. scaling 1080p->720p->1080p is very obviously softer / less detailed then the native 1080p source. keep in mind, this is very atypical. Most sources you can hardly tell a difference.

Obviously with a 720p LCD you wouldn't see the difference at the end of the day, but you might want to plan for the future so you don't have to re-do all your backups at 1080p later ;)

Software players: Media Player Classic - Home Cinema (aka MPC-HC) is the only option for me. End of discussion.

Derek

[mp3dom]

That's total crap. Provide an example of where x264 is worse than any other encoder at any bitrate, not just "high" bitrates.

There's no need to be so upset. Since I'm working with unreleased material I obviusly cannot provide any public 'example' but you can just use other real pro-encoders and see the differences by yourself. BD are not intended to carry a 1080p movie at 5 Mbps (even if someone is doing so on 'professional level' and other are doing so at amateur level) and what I want when I'm encoding a source is to be as close as possible to the master. I'm not saying 'close as possible' during playback but close as possible if I step frame by frame (this exaplain my previous post where I've said that even 40Mbps is not overkill). On animation (not CG) footage with lots of color smoothing, dark areas, subtle grain and other similar footage (example: Evangelion 1.11) I'm more satisfied with the ouput of the pro-encoder than the veryslow preset of x264 at crf 12 (which output more or less the same bitrate but shows more artefacts especially on B frames where color smoothing sometimes is lost, grain retention sometimes shows small macroblocks of no-noise or blocks with different colors and other similar problems). Also the encoding speed of the pro-encoder is even faster so I've no reason to use x264 at high bitrates. It makes sense to me to use x264 at lower bitrates (i.e. extra footage with lower average bitrate) where pro-encoders are weaker and shows in general more problems than x264. I don't understand why, when someone spoke about other pro-encoders, all the 'x264 addicted' automatically comes to denigrate. Is it so impossible to understand that something different to x264 could be sometimes overall better?

I'm interested in the best output quality so I'm not interested in an encoder that need to be 'the best' at any bitrate. If one encoder performs better to me at high bitrates and another works better at middle-low bitrates, I will use both depending the bitrate to use. I'm not saying that x264 is crap but you should also not say that all but x264 is crap. Have you ever used or see the top pro-encoders (CC-HDe is just an example but there are other encoder too) in action to say that x264 is always the 'best' (I'm not referring to MainConcept SDK)? Or you're saying so because others have said that?

Anyway you're 'x264 addicted' so you should not be so upset... just continue to use x264 for everything and live happy.

[AlexW]

@mp3dom
If you can't provide any example clips that show some of the visual problems that you're seeing with x264 at high bitrates it's going to be very difficult for anyone to improve x264 to fix them, since you can't upload any clips of this unreleased material maybe you could try encoding some of the HD clips from http://media.xiph.org/video/derf/ with both the pro encoders and x264 to see if any of them show the problems that you're describing?

[liquidskin76]

Ok, so i loaded up the 4 clips of Avatar (original, cq18, cq20, cq22) in AvsPmod and compared them on a frame-by-frame basis.

The quality shift is clearly visible when comparing down through the different versions. The focus and sharpness drop.

It's amazing you can't see it when viewing at normal speed!

@7ek,

Do you use that avs script for all encodes? Could you post the full script?

Thanks

[shon3i]

I know what mp3dom talking about, and it's true, that Blu-Code and CCE-HDe mostly show more respect at high bitrates to source. I don't say that x264 can't reach transparency, but is sometimes so hard. Most annoying thing that x264 easly produce banding at very high bitrates, while Blu-Code somehow hide it and preserve grain

[Revgen]

^As has been mentioned before, if you can show samples, the x264 developers can take a look and try to improve the issue.

[jpsdr]

Most annoying thing that x264 easly produce banding at very high bitrates.
Does it mean :
At lower bitrate it doesn't produce, but produce at higher, or it's alway producing and even at higher bitrate it's still producing ?
Is it something realy realy bad, or is it "a little" ?
I'll try to check if i notice something on the animes i'm working with now, and eventualy post if i see things, but as my masters are the japanese blu-ray, it may be somehow 'too late', as it's already something encoded.

[7ekno]

@7ek,

Do you use that avs script for all encodes? Could you post the full script?

Nah, I look thru a few random sections of a movie (checking for grain, focus, levels, etc), then decide which scripts to throw at it ...

The example above was a quick example with this simple script:

DGSource(blah)
MCTemporalDenoise(settings="low", gpu=true)

I tend to use some form of "grain reduction" to improve compressability (even if it's a simple Undot() or RemoveGrain(mode=1)) ...

7ek