Codecs: MP4 Format and Saps - Page 3

incredible · #41 12-21-2005, 11:27 AM

Quote:

Originally Posted by Sagittaire

Actually:

2) MPEG4 ASP / LC-AAC / MP4 with chapter/subtitle is supported by SAP (more and more) and you can make that with Nero Recode. You can make anamorphique 720*576 encoding for SAP.

At the german DVDboard where also Productmanagers from Xoro are members, it couldn't be proofed that anamorph mp4 sources do play back properly at the right SAR/PAR. No Mention the other brands which also do mess up in that issue. So the future could bring that feature to more brands/players.

Quote:

3) DivX container is actually the most advanced (for MPEG4) container. You can make integral menu reencoding with exactly the same structure than the original DVD menu and compatibility with SAP will coming.

4) First HD SAP will coming too and you can read full HD MPEG4 ASP encoding
http://forum.surdvd.com/viewtopic.php?t=43651

Even if there do exist 2 or 3 SAPs supporting this .... in the forthcoming month another feature would come up in the mp4 asp/avc development and then again no proper playback support via "these" SAPs will be given.

If you do focus on latest video codec developments and al its features, a HTPC will be obligatory (just imho).

Quote:

In the beginning within the Miramax intro where the camera flys from below to top over a sparkling watersurface I can see blocks. but up to then ... the quality is fantastic.

There are not a MPEG4 problem but MPEG2 source saturation (very high motion with very high complex texture -> very high quant for vbv MPEG2 compliant stream ) ... I have this PAL DVD ...

This wasn't a mpeg4 "neither" a mpeg2 problem. It was a ABR encoding problem where the bitrateallocation is totally different then in 2pass mode.
The Source is seen in its quality very good (for shure in such a complex texture for getting Quants of 2 in mpeg2 a bitrate of 15000 would be needed.

Also the subjective good results out of the h264 compressing technique is NOT only the encoding approach of the codec but also the decoding technique as on such high texture problems the deblocking routines while decoding are far more advanced then at any inner mpeg2 decoder deblocking algorhythm.

Sagittaire · #42 12-21-2005, 05:41 PM

Quote:

At the german DVDboard where also Productmanagers from Xoro are members, it couldn't be proofed that anamorph mp4 sources do play back properly at the right SAR/PAR. No Mention the other brands which also do mess up in that issue. So the future could bring that feature to more brands/players.

Switch simply your SAP is in 16/9 mode. "You can make anamorphique encoding in nero standard for your SAP" is not "anamorph mp4 sources do play back properly at the right SAR/PAR"

In fact most of hold SAP will be able to read mp4 (Nero Standard) or divx (DivX 6 standard) with simple FirmWare Update because SAP are compatible with video codec (MPEG4 ASP) and audio codec (MP3 and LC-AAC). Container, menu, chapter, subtitles are "only" FW problem ...

Quote:

This wasn't a mpeg4 "neither" a mpeg2 problem. It was a ABR encoding problem where the bitrateallocation is totally different then in 2pass mode.
The Source is seen in its quality very good (for shure in such a complex texture for getting Quants of 2 in mpeg2 a bitrate of 15000 would be needed.

MPEG2 in this particular scene (Kill Bill for example) is unable to encode correctly because bitrate RC saturation (for vbv DVD compliance). Encoder must use very high quantizer for vbv compliance.

If you want example see that: I encode this sample with TMPGEnc with your knotch matrix and CQ 100 (9500 Kbps for average bitrate ... !!!)

and it's a blocking festival ...

It is not because MPEG2 is unable to encode correctly this part but only with ultra high local bitrate (parhabs 50 Mbps in q2 for this example). In fact DVD standard is unable to encode very complexe motion + very complexe texture.

rds_correia · #43 12-21-2005, 07:02 PM

EDITED
Please disregard the previous version of this post.
Now I understand why I couldn't download Sagittaire's movie.
The URL is wrong.
Take it here.
Cheers

rds_correia · #44 12-21-2005, 07:23 PM

Ok, I finished watching the source.
To all those that are interested in this, please open lossless.m2v with VirtualDubMod and browse to frame number 259 (which I believe is where Sagittaire got his above posted picture).
May I say that TMPGEnc did a hell of a job by keeping all the dancing blocks that can be seen in the source movie.
Crap in - crap out.
This is not MPEG-2 fault although I am inclined to agree with you that MPEG-2 has his limits and that MPEG-4 has other "weapons" to fight against blocky encodes and against high bitrates.
Cheers

kwag · #45 12-21-2005, 08:02 PM

Quote:

Originally Posted by Sagittaire

Actually:

1) H264 is the most advanced codec in the world.

Try OGG Theora at 720x480 @ around 800Kbps, and compare it to a 720x480 H.264 at that same bitrate.
Then tell me which is more advanced

http://www.v2v.cc/~j/ffmpeg2theora/
However!, the Apple implementation of H.264 is getting better, and it's far superior than the free H.264 implementation floating on the net, which is VERY poor compared to Apple's.

-kwag

kwag · #46 12-21-2005, 08:30 PM

Quote:

Originally Posted by Sagittaire

If you want example see ... I encode this sample with TMPGEnc with your knotch matrix and CQ 100 (9500 Kbps for average bitrate ... !!!)

and it's a blocking festival ...

That's strange, because I just re-encoded the lossless.m2v, without any filtering, and just using our notch matrix, at CQ=85, MAX=9800, MIN=300, and here's the result: http://www.mytempdir.com/335664 (Added bonus, the file is now only 12,264KB. About 2MB smaller )
There's very little difference from the original, and in the end, on the water, there are blocks all over the source too

So as Rui said: Trash-In, Trash-Out (Although the other scenes are very nice)

-kwag

Sagittaire · #47 12-22-2005, 04:30 AM

Quote:

May I say that TMPGEnc did a hell of a job by keeping all the dancing blocks that can be seen in the source movie.
Crap in - crap out.

lol ... you think that I'am noob in encoding ... source is 720p source without blocking. This sequence is particulary blocky just because quant is high for DVD vbv respect.

Be carrefull ... I don't say TMPGEnc and KVCD profil "sucks" ... all the other MPEG2 encoder fails with these extreme sequences ( I test libavcodec, CCE, HC and other). It's just MPEG2 with DVD vbv limitation problem ...

Quote:

Try OGG Theora at 720x480 @ around 800Kbps, and compare it to a 720x480 H.264 at that same bitrate.
Then tell me which is more advanced
http://www.v2v.cc/~j/ffmpeg2theora/

In fact I actually test but x264 HP and ateme HP are very better and by far. Apple H264 implementation is very crap. If you have little vob I can make demonstration with ateme HP encoder. In fact Ateme HP is certainely able to encode 720p at 800 Kbps with good quality ...

Sagittaire · #48 12-22-2005, 05:15 AM

Here just little part of the source
http://multimediacom.free.fr/Download/lossless.avi

Be carefull ... not lossless but MPEG4 ASP quant 1 encoding (in practice it's lossless yor eyes). As you can see not blocking for the source. In fact all this source is incredibily detailled whit very high motion level part (water motion) and with very complexe texturing. This source is a MPEG2 killer ... ideal source for extreme comparison MPEG2 encoder (With this particular source Libavcodec is the best and by far).

And Kill Bill 1 generic introduction is like these particular sequencies. Conclusion: initial problem is not only MPEG4 reencoding but DVD MPEG2 encoding too (If you want screenshoot ... I can make that). "So as Rui said: Trash-In, Trash-Out (Although the other scenes are very nice) " is good analyse in this case ...

incredible · #49 12-22-2005, 06:03 AM

Quote:

Switch simply your SAP is in 16/9 mode. "You can make anamorphique encoding in nero standard for your SAP" is not "anamorph mp4 sources do play back properly at the right SAR/PAR"

In fact most of hold SAP will be able to read mp4 (Nero Standard) or divx (DivX 6 standard) with simple FirmWare Update because SAP are compatible with video codec (MPEG4 ASP) and audio codec (MP3 and LC-AAC). Container, menu, chapter, subtitles are "only" FW problem ...

a) Its the sense of a SAR/PAR Flag to provide to the decoder the information for proper pixelaspectratio final playback on your PC or for correct letterboxing on a 4:3 TV Set. IF the TV Set is 16:9 then a SAP should also have a Widescreen Ouput Mode via OSD so the image will be send to the SAP in a par 128/117 state as it is. The TV finally desqueezes it. If not respecting this, the source height will be decreased and finally width and height will be bumped up "both" on a 16:9 TV.

Till now I've no seen any mpeg4 SAP which does treat a correct PAR/SAR anamorph encoding correct! Even if it reaches the 16:9 TV Set in a correct PAR, you can see the that an internal nonsense interpolation has been done.
Thats also why on even most mpeg4 compatible SAPs no proper mpeg4-interlaced output is given.

b) a simple firmware update isnt the cure on every cases, its also a hardware chipset based issue (D/A converter).

Quote:

initial problem is not only MPEG4 reencoding but DVD MPEG2 encoding too

As I told ...

Quote:

Originally Posted by Inc.

The Source is seen in its quality very good (for shure in such a complex texture for getting Quants of 2 in mpeg2 a bitrate of 15000 would be needed.

"My" KillBill version at the MIRAMAX Intro results quite good in that DVD compilant source state (almost no Blocks, far awy from your "blocks-sample" from that other movie you used), but as said if no Blocks should result a much higher bitrate etc. is needed.

Many People do think DVD "is" the quality but thats a big rumour.
DVD mpeg2 4:2:0 @MainProfile is an end-consumer format.
Hi-Q Source Storages are professionally made at 50Mbit/sec, 4:2:2 and I-frame only mpeg2 (which "almost" equals an mjpeg 4:2:2 stream at 6.25Mb/sec.)

But we where talking about the blocks at the beginning of my "fast'n Trashy-setting ABR x264 encoding" and these where the result due a ABR encoding and its not temporal-based used Bitrateallocation.
And even at these trash-settings I was IMPRESSED about the resulted quality seen in average of the whole encode.

Thanks for the AVI its a good testing source.

Sagittaire · #50 12-22-2005, 06:22 AM

Quote:

Hi-Q Source Storages are professionally made at 50Mbit/sec, 4:2:2 and I-frame only mpeg2

Perhaps not good with my particular source ... I will try ... lol
IMO lossless encoding is always the best way for professional use.

Quote:

a simple firmware update isnt the cure on every cases, its also a hardware chipset based issue (D/A converter).

For *.mp4 and *.divx container it's generally possible with the most commun MPEG4 chip (mediatek, sigma, zoran ... etc) like for example FW nero digital update for old kiss SAP. Don't forget too that .divx container is only a extended .avi container. Large majority of MPEG4 SAP are able to play .divx files correctly without FW update (without xml menu at this time).

incredible · #51 12-22-2005, 10:20 AM

Quote:

Originally Posted by Sagittaire

IMO lossless encoding is always the best way for professional use.

Maybe In your opinion but targets for "professional" use do also lay on professional standard formats like Betacam IMX (Mpeg-2@ 50 MBit/s @ CBR) or Digital Betacam (2:1 DCT Compression) etc.
Not uncompressed raw YUV neither HuffYUV or the common lossless codecs used in the amateur video communities.

Dialhot · #52 12-23-2005, 01:24 PM

For everyone attention :

Quote:

Inutile de discuter avec moi ... j'ai toujours raison ... en tout cas j'en suis convaincus et c'est le principal ... Wink

Means : "no way to discuss with me... I'm always right... at least I'm convinced about that and that's enough".

(avec plein de fautes d'orthographe au passage...)

I let you evaluate the "humor" behind this signature...

rds_correia · #53 12-23-2005, 01:59 PM

LOL good catch.
But up to some extent I think he is right.
MPEG-4 can work pretty well with average or low bitrates whereas MPEG-2 has many limitations in that area.
These limitations are what we try to workaround everyday when we encode in KVCD or KDVD.
Nevertheless I would still like to see an encoder (MPEG1/2/4) handle is lossless.mpg and make the encoded clip better than the source clip.
TMPGEnc did a hell of a job on his baaaaaad source

.
Cheers

kwag · #54 12-23-2005, 06:57 PM

Quote:

Originally Posted by rds_correia

MPEG-4 can work pretty well with average or low bitrates whereas MPEG-2 has many limitations in that area.

For sure, BUT MPEG-2 scales FAR better than MP4 or even H.264 on very high bitrates, and that's a fact

That's why MPEG-2, I frame only @50Mbps, is what's used in broadcast studios for high quality archival

Edit: BTW, a friend of mine does just that, and has done several commercials for customers who air on "***vision", and guess what format he delivers the material

(But nobody is aware that KDVD Notch is in there

)
Sorry, but I'm not allowed to give out more information

-kwag

Dialhot · #55 12-23-2005, 07:16 PM

Is it a new commercial usage of the Notch ?

kwag · #56 12-23-2005, 08:19 PM

Quote:

Originally Posted by Dialhot

Is it a new commercial usage of the Notch ?

Actually, it's been used for over a year

Just take TMPEG and encode something MIN=300, MAX=50,000, and use the Notch matrix and I frame only, CQ=100.
See what you get

Also, I'll let you all in on a little secret, because I think it's time

On high average bitrates, KDVD Notch PRODUCES CLEANER VIDEO and less macroblocks than the industry standard MPEG matrix, and I will now provide a clear example of what I mean.

Using a piece of the clip that Sagittaire provided, here's a comparison of an encode using TMPEG with "standard" MPEG matrix and KDVD Notch.
Both encodes were made with MIN=300, MAX=50,000, and CQ=40 using *I frame Only*.

Notch Matrix:

Standard MPEG Matrix:

I think the benefits of the Notch don't need an explanation

This is perfect for HD use, standard broadcast resolutions, or MPEG-2 video archival.

-kwag

Sagittaire · #57 12-25-2005, 11:40 AM

Quote:

For sure, BUT MPEG-2 scales FAR better than MP4 or even H.264 on very high bitrates, and that's a fact
That's why MPEG-2, I frame only @50Mbps, is what's used in broadcast studios for high quality archival

1) Perhaps because industry are simply not ready for that ... IMO good H264 encoder is the best in all situation but it's true for low quantizer differencies between H264 and MPEG2 are small.

2) For very high bitrate you can use lossless mode for H264. Industry use 50 Mbps stream. H264 LossLess mode done 25 Mbps for this stream ... IMO H264 lossless mode is better than MPEG2 lossy mode ...

3) I will forget : initial source is 1280*720 H264 at only 5 Mbps ... and make with the very crap QT7 AVC H264 encoder ...

Quote:

no way to discuss with me... I'm always right... at least I'm convinced about that and that's enough

good signature, isn't it ...

Quote:

TMPGEnc did a hell of a job on his baaaaaad source

all the MPEG2 encoder make bad job for the complete source.