why blue ray for h264 hd films???

[inurenegade]

what about the factor of the new high def audio?
arent they doing like 192khz now and 7.1 and isnt there like uncompressed audio for the highest fidelity?
imagine the space that must take up on a DVD9

[Sagittaire]

what about the factor of the new high def audio?
arent they doing like 192khz now and 7.1 and isnt there like uncompressed audio for the highest fidelity?
imagine the space that must take up on a DVD9

Actualy HDDVD use Dolby Digital Plus at 640 Kbps for multiple audio stream 7.1 48 Khz ...

8500 MB HDDVD structure on DVD9 for 120 min movie
simple authoring with menu, chapters, subtitles, languages : ~500 MB
Multiple (3 for exemple) DD+ audio stream at 640 Kbps (7.1 and 48 Khz) : ~1750 MB
1080p Video stream at 7 Mbps : ~6250 MB



Like I say extra storage from HDDVD optical support is only for complexe authoring with extra high quality:
- PIP video stream bonus
- LossLess audio stream
- extra high quality for video stream

22500 MB HDDVD structure on HDDVD30 for 120 min movie
complexe authoring with menu, chapters, subtitles, languages : ~1000 MB
Unique principal audio stream DD trueHD at 3 Mbps: ~ 2750 MB
Multiple secondary (3 for exemple) DD+ audio stream at 640 Kbps (7.1 and 48 Khz) : ~1750 MB
Principal Video stream 1080p at 15 Mbps : ~13500 MB
Secondary Video stream 480p at 4 Mbps : ~3500 MB

[Morte66]

If you want real comparison then quality for 1080p 120 min movie on DVD9 will be better than QT AVC trailers.

I agree. I've done some encodes from uncompressed 1080p with fairly difficult material. At around 4500kbps they look better than the QT trailers and radically better than a 25000kbps MPEG2 encode. They're excellent at 8000kbps with x264. Maybe 12000kbps to top out under critical examination on my fairly brutal monitor.

Settings for x264 were slow but not silly:

--crf 22 [or whatever] --ref 3 --mixed-refs --no-fast-pskip --bframes 3 --b-pyramid --b-rdo --bime --weightb --direct none --filter -2,-2 --subme 6 --trellis 1 --analyse all --8x8dct --threads 2 --thread-input --cqmfile "C:\Program Files\MeGUI\extra\M4G-V3.cfg"

It seems that the cleanliness of the source has a heck of a lot to do with it. Re-encoding 1080 MPEG2 transport streams in constant quality mode after some avisynth work, I typically come out about 10GB for a movie. On a really excellent clean broadcast of "Million Dollar Baby", I got 4.2GB. A crappy, noisy stream of "Serenity" was heading for 16GB, before I realised that the fractal denoiser was just revealing the blocking and killed it. So studios, working from masters, could make very good encodes at DVD9 bitrates if they wanted to. But bigger discs would allow better quality for people with high-end displays.

There is perhaps a catch: I can't play these encodes on my A64 X2 3800+ unless I'm using CoreAVC Pro. No other decoder is fast enough. The VC1 encodes they're putting on HD DVD are probably a bit easier.

In the final analysis, I think skill and care in the capture/encoding process make more difference than 6kbps vs 12kbps for x264.

[travisbell]

At around 4500kbps they look better than the QT trailers and radically better than a 25kbps MPEG2 encode.

I am guessing you mean 25mbps...

I would love to see some of the examples... any chance you could post them? I would especially really like to see the 1080p encode from your uncompressed source... sounds like it should tickle my eyes pretty good!

[Morte66]

I am guessing you mean 25mbps...

I would love to see some of the examples... any chance you could post them? I would especially really like to see the 1080p encode from your uncompressed source... sounds like it should tickle my eyes pretty good!

Yep, 25mbps, edited.

I'll see if I can dig them out tomorrow.

[reepa]

If you need a "film master" for testing different codecs, you can get SVT test sequences (thank you Swedes!) from:
ftp://vqeg.its.bldrdoc.gov/HDTV/SVT_MultiFormat/

They're meant for testing television broadcasts, which is why they are 50fps. Each frame is 3840 by 2160 (four times 1080p!) and 48 bits per color (16-bit per channel vs 8-bit per channel!). The sequences were captured on 65mm film for the absolute best quality.

These sequences are ideal for testing since they have real film grain (compared to the Elephants Dream sequences). I haven't done any testing because I don't have the hard drive space required (a single frame is almost 50 megabytes!) nor the expertise in video compression.

[Morte66]

I would love to see some of the examples... any chance you could post them? I would especially really like to see the 1080p encode from your uncompressed source...

http://rapidshare.com/files/7325856/Taurus.x264.7314kbps.deblock-2-2.cqm-M4GV3.mp4.html

99MB encode from 8.24GB of uncompressed digital camera footage, 113 seconds at 25fps, 7314kbps, x264 reports final ratefactor=23.36.

Here are the notes on the source footage:

Sequence #Frames Short description
--------------------------------------------------
Blue sky 250 Top of two trees against blue sky. High contrast, small color differences in the sky, many details. Camera rotation.
Pedestrian Area 375 Shot of a pedestrian area. Low camera position, people pass by very close to the camera. High depth of field. Static camera.
Rush-hour 500 Rush-hour in Munich city. Many cars moving slowly, high depth of focus. Fixed camera.
Riverbed 250 Riverbed seen through the water. Very hard to code.
Station 313 View from a bridge to munich station. Evening shot. Long zoom out. Many details, regular structures (tracks)
Sunflower 500 Sunflower, very detailed shot. One bee at the sunflower, small color differences and very bright yellow. Fixed camera, small global motion.
Tractor 761 A tractor in a field. Whole sequence contains parts that are very zoomed in and a total view. Camera is following the tractor, chaotic object movement, structure of a harvested field. Very red wheels of the tractor



Camera: Sony HDW-F900
Recorded on (Tape): HDCam
Stored on: DVS
Frame rate: 25 fps (progressive)
Resolution: 1920x1080
Color subsampling: 4:2:0
Filter Tabs for Subsampling: -0.0063 / 0 / 0.0299 / 0 / -0.0831 / 0 / 0.3098 / 0.4994 / 0.3098 / 0 / -0.0831 / 0 / 0.0299 / 0 / -0.0063
Color conversion: ITU Rec BT 709 (SMPTE 274M)
Original files contact: oelbaum@ei.tum.de
Restrictions of use: No restrictions
Copyright: No Copyright
Date of Recording: Summer 2001
Source: Taurus Media Technik, Dr. Karl Mauthe
Producer: Martin Kreitl martin.kreitl@KirchGruppe.de
Camera Operator: Jürgen Würzinger
Camera Assistent: Yean Ives Diss

All material was recorded in summer 2001 by Taurus Media Technik.

The x264 command line for pass 2 was: --pass 2 --bitrate 7295 --stats "D:\MeGUI\Samples\samples.stats" --ref 3 --mixed-refs --no-fast-pskip --bframes 3 --b-pyramid --b-rdo --bime --weightb --direct none --filter -2,-2 --subme 6 --trellis 1 --analyse all --8x8dct --threads 2 --thread-input --cqmfile "C:\Program Files\MeGUI\extra\M4G-V3.cfg"

sounds like it should tickle my eyes pretty good!

My first thought on seeing the original was "that's really soft", maybe because it's devoid of all the grain/noise/sharpening we're used to on DVD... If you look carefully and find the point of focus (e.g. the bee not the sunflower) it's actually pretty sharp.

The encode looks close to the original in general quality. No mp4 encoder I used gets the clouds in the first section completely free of banding; Xvid at q=2 (about 25mbps) comes closest to avoiding banding but it has other faults.

As for HD DVD vs DVD9... At about 14 seconds there's a guy who walks into the frame from the right wearing red dungarees. If you look at his back on a good display you see a little blocking. Add 50% to the bitrate and that goes away. I suspect you'd need to bump it again to cover film grain and sharpening.

[giandrea]

I'm downloading, will have a look at the results, even if at that ratefactor they should be more than good for the casual observer.
But the question here is if this encode is compatible with BlueRay or HD DVD specs. If it is not, the whole point of your post is, well, pointless... :P

[SeeMoreDigital]

Thanks for the sample Morte66.

It looks beautiful when viewed (at the correct distance) on my 42" high-def screen...

A very nice example of what can be achieved at below 8Mbps ;)


Cheers

[drmpeg]

Same content in MPEG-2, 18 Mbps average, 30 Mbps peak in a 35 Mbps TS.

http://www.w6rz.net/1080p25.zip

259 MB.

Ron

[Egladil]

I don't get it. Somebody here using crf 18, another crf 22 -I consider that LQ (well 18 is not LQ maybe, but not HQ), they'll be eating details. The point is that with HDDVD there's enough space to store the video in nearly transparent quality (I did some tests and for me crf 15 looks good - and crf15 needs a lot of bitrate) - I'd never buy a DVD9 with overcompressed (and possibly denoised, ...) movie. I mean thats fine for somebody or maybe for personal backups, but the whole point of hddvd is not compression, but quality. If i'd buy that I want to be sure that the movie looks exactly like in cinema - if it's grainy (james ryan for example) - then it has to be grainy on a hddvd, and not denoised so that it can be compressed better.

[Morte66]

But the question here is if this encode is compatible with BlueRay or HD DVD specs. If it is not, the whole point of your post is, well, pointless... :P

Um, *shrug*. If there's anything compatible about it, it's accidental.

I thought the issue here was "would we really need larger discs if we used the best current encoders and techniques on DVD9?". And I think the answer is "not for most people, though serious enthusiasts would see improvements on a bigger disc".

[Morte66]

I don't get it. Somebody here using crf 18, another crf 22 -I consider that LQ (well 18 is not LQ maybe, but not HQ), they'll be eating details. The point is that with HDDVD there's enough space to store the video in nearly transparent quality (I did some tests and for me crf 15 looks good - and crf15 needs a lot of bitrate) - I'd never buy a DVD9 with overcompressed (and possibly denoised, ...) movie. I mean thats fine for somebody or maybe for personal backups, but the whole point of hddvd is not compression, but quality. If i'd buy that I want to be sure that the movie looks exactly like in cinema - if it's grainy (james ryan for example) - then it has to be grainy on a hddvd, and not denoised so that it can be compressed better.

Well, OK. That's pretty much what I've been driving at -- DVD9 is fine for many/most people provided you use the best encoding available, but not for everybody. You're the "not everybody" part of the test group. ;)

A semi-tangent: we've discussed HD DVD and Blu-Ray almost interchangeably in this thread, but they're different. Blu-Ray a bigger but is a lot more expensive to manufacture, the plants are a bigger investment, and the drives are costlier. And it's superfluous, because HD DVD is enough for everybody if you use a good encoder. Blu-Ray only has a use of sorts with MPEG2, and MPEG2 on Blu-Ray looks way worse than h264 on DVD9.

[DarkZell666]

Actually, crf18 is near-transparent already on SD, and it's even more transparent on HD material. What you seem to call details is pixel-precise information, which you won't have by the simple fact of using something else than a lossless codec.

On my PC screen, the 1920x1080 picture get's downsized to 1024x576, which is nearly half the resolution (and nearly 4x less pixels). Considering that AVC offers a minimum block size of 4x4, this block once resized is only represented by 2x2 block (4x less pixels). So any pixel-level information becomes subpixel information: visually negligible.

This spawns another interesting question (at least I don't have a clue about the info, maybe someone else has): do HD screens REALLY have 1920x1080 dots on them ? At least computer screens aren't really there yet (which is probably why I can't spot a difference right now).

But visually, the difference between crf15 and crf18, is (imho) not obvious at all (unless you substract the 2 encodes with each other and look at the difference). It's just like comparing 320kbps CBR mp3 and 256kbps VBR mp3. You can't hear the difference but you "know" it's different. Sort of a placebo effect if you want my opinion :p

As for the grain stuff, x264 (by default) doesn't keep much grain, but you can tweak CQM and deblocking stuff to keep it. Nero manages not too bad on grain because it has relatively good psy-enhancements. Also note that in HD content, the grain will be much fatter (one "grain" will occupy more pixels to look the same), and won't be processed as hard as in SD material. Grain will look much more like spatial information (instead of noise) to the codec.

Well, this is just my four cents. There's another crystal-clear source we could try: http://orange.blender.org/ (Elephant Dream). The unprocessed video and audio is downloadable here: http://media.xiph.org/ED/ (the info is from another thread in the General discussion forum).

[SeeMoreDigital]

This spawns another interesting question (at least I don't have a clue about the info, maybe someone else has): do HD screens REALLY have 1920x1080 dots on them ? At least computer screens aren't really there yet (which is probably why I can't spot a difference right now). Full high-def "progressive" LCD screens offering a native resolution of 1920x1080 (2,073,600 total) pixels actually have 3 times this amount of pixels - one for each primary colour.


Cheers

[Morte66]

transparent

...is in the eye of the beholder. It depends on how fussy you are, how hard you're looking, what equipment you've got, and your viewing conditions. An encode that was just transparent on my 8 year old CRT monitor wouldn't be on my new 1920x1200 LCD.

It's just like comparing 320kbps CBR mp3 and 256kbps VBR mp3. You can't hear the difference but you "know" it's different. Sort of a placebo effect if you want my opinion :p

Well, I can tell that difference, but then I'm a hi-fi nut with a seventy thousand pound audio system. My brother can't tell 192k MP3 from the WAV file. The interesting thing for me is that I don't really care about 256vs320 MP3. They both convey the musically important elements equally well, the difference is sonic rather than musical. But I do prefer AAC/OGG 192 to any MP3 -- MP3 just can't do cymbals right.

I'm less fussy about video than audio: I see the difference between crf22/18/16 if I want to, but I'm not really bothered about it.

[SeeMoreDigital]

But I do prefer AAC/OGG 192 to any MP3 -- MP3 just can't do cymbals right. Most probably due to MPEG-1 Layer-3 audio not being able to offer a wide enough dynamic range over the mid-to-high frequency bands....

[Morte66]

If you need a "film master" for testing different codecs, you can get SVT test sequences (thank you Swedes!) from:
ftp://vqeg.its.bldrdoc.gov/HDTV/SVT_MultiFormat/

They're meant for testing television broadcasts, which is why they are 50fps. Each frame is 3840 by 2160 (four times 1080p!) and 48 bits per color (16-bit per channel vs 8-bit per channel!). The sequences were captured on 65mm film for the absolute best quality.

Gulp. That's going to be some download.

And you really need the 2160p versions, to get realistic grain aliasing in the scans. [A 2160p film scan resized to 1080p looks smoother than a 1080p scan. Something I learned the hard way scanning still photos.] Oh well, here goes...

[Morte66]

Same content in MPEG-2, 18 Mbps average, 30 Mbps peak in a 35 Mbps TS.

http://www.w6rz.net/1080p25.zip

259 MB.

Nice encode, better than I managed with MPEG2. How did you encode it?

[smok3]

This spawns another interesting question (at least I don't have a clue about the info, maybe someone else has): do HD screens REALLY have 1920x1080 dots on them ? At least computer screens aren't really there yet (which is probably why I can't spot a difference right now).

my samsung runs at 1920x1200 right now, so what do you mean? (or should i count the pixels manually? :))

[DarkZell666]

...is in the eye of the beholder.Agreed, lol. But I'm really having a hard time understanding how people can possibly see such differences ... o_O I'll have to reconsider my way of (litterally speaking) "seeing" things ;)

Edit for smok3: grrr XD, the max I've ever seen is the 1600*1200 LCD I have at work, and "average joe" buying his computer at the local supermarket will be given a 1280*x CRT most of the time. Some of you guys really have uncommon or expensive hardware :p

[SeeMoreDigital]

Same content in MPEG-2, 18 Mbps average, 30 Mbps peak in a 35 Mbps TS.Looks good Ron.... I can play these 1920x1080 MPEG-2 samples in hardware ;)

[akupenguin]

And you really need the 2160p versions, to get realistic grain aliasing in the scans. [A 2160p film scan resized to 1080p looks smoother than a 1080p scan. Something I learned the hard way scanning still photos.]
True. But do you know whether most HD movies are scanned at only 1080p, rather than downscaled?

[Morte66]

True. But do you know whether most HD movies are scanned at only 1080p, rather than downscaled?

Come to think of it, I have absolutely no idea.

Damn.

I wonder if I can find out...

[reepa]

Gulp. That's going to be some download.

And you really need the 2160p versions, to get realistic grain aliasing in the scans. [A 2160p film scan resized to 1080p looks smoother than a 1080p scan. Something I learned the hard way scanning still photos.] Oh well, here goes...

The lower resolution versions have been downsampled from the master 2160p scans so in theory you'll lose nothing by downloading the 1080p versions (unless you want to use a different resampling algorithm, SVT used Lanczos).

[DarkZell666]

my samsung runs at 1920x1200 right now, so what do you mean? (or should i count the pixels manually? )After re-reading this a couple of times, I just understood what you meant :o In fact, I believe some screens don't have physically the number of pixels available on the screen to display the chosen resolution. Instead, the screen does some resizing/interpolating/downsampling or whatever to get all the video info on the screen. So my answer to your question is in fact indeed: yes, count them manually xD (or go read the manual :p)

I say this because on LCD's, there's always a particular resolution where the pixels fall correctly into place, and all the others resolutions look wierd and pixelated (just like if you were watching super-nintendo graphics). That resolution where the pixels fall into place is physically the number of dots on the screen (at least that's what I concluded from what I've seen). (read here:http://en.wikipedia.org/wiki/Display_resolution#Current_standards_in_resolution)

[SeeMoreDigital]

Looks good Ron.... I can play these 1920x1080 MPEG-2 samples in hardware ;)Actually, after closer inspection, it seems as though the people walking about in the second scene are suffering from something akin to a field blending effect :scared:

[smok3]

True. But do you know whether most HD movies are scanned at only 1080p, rather than downscaled?

i would say that new movies are probably simply digital intermediates downscaled, older movies are probably scanned at exact resolution. (at least that could be a good guess)

[Morte66]

Actually, after closer inspection, it seems as though the people walking about in the second scene are suffering from something akin to a field blending effect :scared:
If you're thinking of what I'm thinking of, that's a combination of 24fps video, a shutter shorter than 1/24 second, motion, and image persistence in your retina/brain creating an "echo". If you play the material at 5fps it vanishes. At 25fps, it's there in the original too.

[SeeMoreDigital]

If you're thinking of what I'm thinking of, that's a combination of 24fps video, a shutter shorter than 1/24 second, motion, and image persistence in your retina/brain creating an "echo". If you play the material at 5fps it vanishes. At 25fps, it's there in the original too.Hmmm!

I've seen this effect loads of times, when people try to generate progressive encodes from pure interlaced sources!

I think we need to know more about the primary source ;)

[Morte66]

The lower resolution versions have been downsampled from the master 2160p scans so in theory you'll lose nothing by downloading the 1080p versions (unless you want to use a different resampling algorithm, SVT used Lanczos).

It's not quite so simple, for two reasons...

1. When you scan film with grain (silver-based) or dye clumps (chromogenic), you get sample aliasing. It's especially noticable when the pixel size in your scanner is somewhat similar to the grain/clump size on the film. A 2160p scan could create far more or far less of this noise than a 1080p scan, depending on the exact dimensions involved. It's not especially predictable or intuitive. [Practical example: Fuji NPH400 looks fine scanned at 1800dpi or 2900dpi, but noisy as hell at 2400dpi.]

2. When you resize a bitmap containing random noise down, you lose some of the noise through signal averaging. If it's perfectly random you'll get the sqaure root of your data reduction (so reducing pixels by a factor of four would reduce your noise by a factor of two), but it's never that simple in practice.

[Morte66]

i would say that new movies are probably simply digital intermediates downscaled, older movies are probably scanned at exact resolution. (at least that could be a good guess)

That's consistent with what a couple of people have told me this morning.

On a more practical level, the 2160p files add up to about 90GB and I've only got 70GB free...

[Morte66]

Hmmm!

I've seen this effect loads of times, when people try to generate progressive encodes from pure interlaced sources!

That's exactly what I thought when I first saw it. Then I thought maybe it was some sort of ringing/echoes in the encoder. Eventually I discovered that it was echoes in my brain.

I think we need to know more about the primary source ;)

ftp://ftp.ldv.e-technik.tu-muenchen.de/pub/test_sequences/

Took me about 3 days to download. You need the RawSource filter for AviSynth and something like:
v1=RawSource("blue_sky.yuv",1920,1080,"I420")
v2=RawSource("pedestrian_area.yuv",1920,1080,"I420")
v3=RawSource("riverbed.yuv",1920,1080,"I420")
v4=RawSource("rush_hour.yuv",1920,1080,"I420")
v5=RawSource("station2.yuv",1920,1080,"I420")
v6=RawSource("sunflower.yuv",1920,1080,"I420")
v7=RawSource("tractor.yuv",1920,1080,"I420")

v1+v2+v3+v4+v5+v6+v7

AssumeFPS(25)

[SeeMoreDigital]

That's exactly what I thought when I first saw it. Then I thought maybe it was some sort of ringing/echoes in the encoder. Eventually I discovered that it was echoes in my brain. When you capture a single frame (from either the x264 or MPEG-2 encode) there's motion missing and blurring evident

x264 encode: -
http://img431.imageshack.us/img431/1869/x264jo1.th.jpg (http://img431.imageshack.us/my.php?image=x264jo1.jpg)

MPEG-2 encode: -
http://img281.imageshack.us/img281/5294/mpeg2re8.th.jpg (http://img281.imageshack.us/my.php?image=mpeg2re8.jpg)

The poor old guy with the ruck-sack has lost his leg below his knee.... I wonder if he knows!

[Morte66]

When you capture a single frame (from either the x264 or MPEG-2 encode) there's motion missing and blurring evident

Sure, the shutter is like 1/50 or 1/100 second or something. Read any primer for still photography, and it'll tell you that you need 1/500 for "reasonable sharpness" with people walking across the frame. That's advice aimed at people who mostly make 6x4 inch prints.

And there are gaps between frames, where the shutter was closed and nothing was captured.

Welcome to digital video standards designed for compatibility with hundred year old cinema technology...

[SeeMoreDigital]

If you still have the "pedestrian_area.YUV" source... What does the original frame look like?

[Morte66]

If you still have the "pedestrian_area.YUV" source... What does the original frame look like?

http://www.joel-benford.co.uk/posts/snapshot20061215122515.jpg

[SeeMoreDigital]

Poor bloke :(

[Morte66]

@SeeMoreDigital...

Here's a partial quote that may interest you, by a poster called RTFM on the AVS forums:
Sometimes watching a film a cinema does a serious number on me with motion judder at certain camera panning speeds.

When I was working as a camera operator I payed close attention to my "Bible" which was the American Cinematographer Manual which has a section titled "35mm Camera Panning Speed Recommendations" which is all about avoiding the "skipping" effect as they described it.

It goes into great detail as to why we perceive this side effect and I quote: "During the panning of the camera in a static setting, a certain displacement of objects takes place on the film from frame to frame. When the film is viewed in projection, this displacement, as it appears on the screen, also constitutes a displacement on the retina of the observer's eye. The viewing cells of the retina, however, are not directly adjacent to each other-- they are spaced at approximately 7.14 minutes of the arc. If therefore, an object viewed on the screen is displaced more than can be sensed within the above angle on the retina, it causes skipping of the viewing cells, which in turn disrupts or chops the continuity of the panning motion"

This would also apply to an object moving across the frame at certain speeds in a static camera setup.

Here's an example from the tables in the ACM: 24frames per second,180 degree shutter, 50 mm lens. A 90 degree pan should take 23 seconds to avoid judder.

Early wide-screen movies shot in 65/70mm such as Oklahoma and Around The World In 80 Days were shot at 30fps rather than 24 to smooth panning shots and cross-screen action.

I presume that when they shoot a Hollywood movie, they're careful to avoid/limit this sort of thing. But the Taurus Media footage, which is specifically meant to be difficult samples for encoder testing, let it happen.

[SeeMoreDigital]

I actually have a few friends dotted around the world, who are professional film camera-men (DOP's). All are having to "re-learn" their craft as the art of capturing images moves away from film and into the digital domain!

That said, having a camera "locked down" and capturing people moving across a frame at "walking speed" should be a minimum requirement for any digital camera...

[Blue_MiSfit]

http://store.apple.com/1-800-MY-APPLE/WebObjects/AppleStore?family=AppleDisplays

Very high resolution display.

Almost native 4k.

Still, it's a solid 4 megapixels + :D

[Sagittaire]

Make comparison between MPEG2 and H264 for HDDVD is difficult. H264 can use HP@L4.1 with specific vbv limitation and MPEG2 can use MP@HL with compliant vbv limitation.

Actualy in most case principal HD video stream for HDDVD use these setting:
- Max GOP length at 0.6006 sec
- 2 adaptative bframe
- max bitrate at ~20 Mbps and buffer at 9781, 14745 and 30000 Kbits for MPEG2, VC1 and H264
- Max Vertical Motion Vector length at 128, 256 and 512 pixels for MPEG2, VC1 and H264

Make comparison between unconstrained crf 15 or crf 20 is completely useless without vbv limitation and GOP restriction. IMO with same vbv crf 15 and crf 20 will done the same local bitrate in difficult part and exactly the same artefacts.

[oddball]

The movies that get away with compression most are of course CG or all digital transfers. I have a couple of CG videos on DVD5 with full 1080P res and they look great. Also certain well known movies that were shot all digital that looks great too. However I also have some rather grainy movies that even if given the best bitrate possible would still look rather dicey. Even recent flicks can have inherent amounts of film grain in them that causes compression of any sort a problem. The odd thing is that even in movies that are shot all digital they deliberately introduce artificial film grain the make a movie more 'cinematic' and to avoid looing somewhat unnatural when shown on a cinema size screen.

[giandrea]

Make comparison between unconstrained crf 15 or crf 20 is completely useless without vbv limitation and GOP restriction. IMO with same vbv crf 15 and crf 20 will done the same local bitrate in difficult part and exactly the same artefacts.

It would be interesting to try again the test with these constrains. Do you know what command line options are required to compress an HD DVD or Blue Ray compliant H264 stream with x264?

If the quality is acceptable it would be possible to make full resolution backups of an HD DVD or BD on a DVD9. :thanks:

[Sharktooth]

Specs are not publically available.

[Sagittaire]

It would be interesting to try again the test with these constrains. Do you know what command line options are required to compress an HD DVD or Blue Ray compliant H264 stream with x264?


For best possible quality with x264:
x264.exe --keyint 15 --min-keyint 1 --vbv-maxrate 28000 --vbv-bufsize 30000 --mvrange 512 --level 4.1 --bframe 2 --b-rdo --bime --weightb --ref 3 --mixed-refs --direct auto --deblock -1:-1 --crf 20 --qcomp 0.75 --ipratio 1.10 --pbratio 1.30 --partitions "all" --8x8dct --me "umh" --subme 7 --no-fast-pskip --no-dct-decimate --trellis 2 --progress -o H264_6Mbps.mp4 HDDVD.avs

and here IMO the best possible quality/speed setting:
x264.exe --keyint 15 --min-keyint 1 --vbv-maxrate 28000 --vbv-bufsize 30000 --mvrange 512 --level 4.1 --bframe 2 --b-rdo --bime --weightb --ref 3 --mixed-refs --direct auto --deblock -1:-1 --crf 20 --qcomp 0.75 --ipratio 1.10 --pbratio 1.30 --partitions "all" --8x8dct --me "hex" --subme 6 --no-fast-pskip --no-dct-decimate --trellis 1 --progress -o H264_6Mbps.mp4 HDDVD.avs


If the quality is acceptable it would be possible to make full resolution backups of an HD DVD or BD on a DVD9. :thanks:

No doubt for that. H264 is able to obtain very high quality in 6-8 Mbps bitrate range for 1080p source and particulary for 2.35 movie.

For same source use 7.0 Mbps for H264 encoding in 1080p25 will done the same quality by pixel than 4.0 Mbps for MPEG2 encoding in 576p25.

[giandrea]

For best possible quality with x264:
x264.exe --keyint 15 --min-keyint 1 --vbv-maxrate 28000 --vbv-bufsize 30000 --mvrange 512 --level 4.1 --bframe 2 --b-rdo --bime --weightb --ref 3 --mixed-refs --direct auto --deblock -1:-1 --crf 20 --qcomp 0.75 --ipratio 1.10 --pbratio 1.30 --partitions "all" --8x8dct --me "umh" --subme 7 --no-fast-pskip --no-dct-decimate --trellis 2 --progress -o H264_6Mbps.mp4 HDDVD.avs

and here IMO the best possible quality/speed setting:
x264.exe --keyint 15 --min-keyint 1 --vbv-maxrate 28000 --vbv-bufsize 30000 --mvrange 512 --level 4.1 --bframe 2 --b-rdo --bime --weightb --ref 3 --mixed-refs --direct auto --deblock -1:-1 --crf 20 --qcomp 0.75 --ipratio 1.10 --pbratio 1.30 --partitions "all" --8x8dct --me "hex" --subme 6 --no-fast-pskip --no-dct-decimate --trellis 1 --progress -o H264_6Mbps.mp4 HDDVD.avs




No doubt for that. H264 is able to obtain very high quality in 6-8 Mbps bitrate range for 1080p source and particulary for 2.35 movie.

For same source use 7.0 Mbps for H264 encoding in 1080p25 will done the same quality by pixel than 4.0 Mbps for MPEG2 encoding in 576p25.

Very interesting, thanks. :goodpost:

So, a bit off topic... would I be able to play a video encoded to be HD DVD or BD compatible, muxed the MP4 container on a PlayStation 3? And anyway I should be able to play it back if burned on a DVD with the BD file structure... right?
Thanks again! :)

[smok3]

The odd thing is that even in movies that are shot all digital they deliberately introduce artificial film grain the make a movie more 'cinematic' and to avoid looing somewhat unnatural when shown on a cinema size screen. yes, 'magic bullet' and packages like that are really popular as it seems.

[Morte66]

If the quality is acceptable it would be possible to make full resolution backups of an HD DVD or BD on a DVD9. :thanks:

I'm sure you could do that and get satisfying quality, but two-pass 1080p encodes at that sort of quality would take several days on a dual core PC. That's an awful lot of encoding for one backup. If possible (tools, DRM) I'd be more inclined to just split the thing over a few DVDs.

I sometimes do 1080p encodes from MPEG2 HDTV so I can use avisynth on the lousy video, that lets me get a backup that's better than the original. But I don't see much point with HD DVD.

[smok3]

from a 'small' user point of view the new formats are especially interesting for data backups imho (especially if you deal with your own production video) - as soon as the price drops..., uhmm, hopefully meantime we can expect a drop price for dual DVDs? :)