Codecs: MP4 Format and Saps - Page 6

digitall.doc · #**101** 01-10-2006, 02:56 AM

Quote:

Originally Posted by Sagittaire

you can post little example with source ...

No, I cannot. The tests were useless, they didn't play well in my (any?) SAP. I usually try to keep as less trash in my hard drive as I can. So I deleted them.
But I saw bitrate spikes in Bitrate Viewer, and maybe more important, my SAP didn't play well. Don't know if it was related to spikes, or VBV under/overflows, but it didn't work.
But even without example, you may trust me... will you?. It didn't play in my SAP. And I was really interested in it to work... but it didn't.

Quote:

Originally Posted by Sagittaire

yes ... not really practical and easy for quick check ... Wink

Well, it is very important (to me it's esential) that stream can be played in almost any SAP. I can't figure other way to see if it will be playable... but playing it. What other method do you suggest?.

Sagittaire · #**102** 01-10-2006, 05:23 AM

Quote:

Originally Posted by digitall.doc

Well, it is very important (to me it's esential) that stream can be played in almost any SAP. I can't figure other way to see if it will be playable... but playing it. What other method do you suggest?.

well it's just question ...

But DivXNetwork propose DMFCheck.exe for new .divx container. DMFCheck check all the video stream, audio stream, menu xml, sub xsub in .divx files and return info for DivX Ultra Profil compliance. Perhaps equivalent DVDCheck.exe exist for m2v files or vob+ifo structure ... ???

incredible · #**103** 01-10-2006, 06:00 AM

Quote:

But little question : how do you check MPEG2 DVD compliant stream ... ???

Do I see that right? You are even not really shure about the compilancy as you dont know how to "really" proof if a given mpeg2 stream is compilant under all circumstancies? But you forced your statement in that "other" tread obout that mencoder mpeg2 outputs are full mpeg2@main profile DVD SAP compilant.

I did post in that thread that only many tests using diff. SAPs can be seen as a "proof". And thats why some few people in here do clearly say that its NOT fully compilant as they have problems while playing back them on "their" special SAP model - and these people know how to encode a mpeg2 stream for compilant DVD purposes. BTW. that problem is ALSO mentioned in the Mencoder docs as Dialhot already quoted these docs parts.

Thats why we said that the time showed that there where no Topics in forums where CCE, TmpgEnc, MCE etc. did fail - well at least i dont know any topic on these. But Mencoder.exe did bring some topics up based on exactly "that" issue.
EVEN the HCenc brought a topic up where a person on the one hand had no props playing back a stream out of tmpegEnc but on HCenc.

Do encode what you want, using what you want if you do encodes for "yourself" then ist no problem.
Do let decide your eyes not maths.

Sagittaire · #**104** 01-10-2006, 07:31 AM

Quote:

Do encode what you want, using what you want if you do encodes for "yourself" then ist no problem.
Do let decide your eyes not maths.

But compliance problem is not eyes problem ... it's only "maths" probleme (me_range, vbv respect, max bitrate, GOP, size ...). DVD compliant chip can or can't read stream ...

Quote:

I did post in that thread that only many tests using diff. SAPs can be seen as a "proof". And thats why some few people in here do clearly say that its NOT fully compilant as they have problems while playing back them on "their" special SAP model - and these people know how to encode a mpeg2 stream for compilant DVD purposes. BTW. that problem is ALSO mentioned in the Mencoder docs as Dialhot already quoted these docs parts.

In fact Mencoder seem able for my 3 SAP to encode without problem (overflow or other) but perhaps too that my 3 SAP are "overcompliancy" .

Possible that somebody shows specific example with accessible source for its SAP ... ??? I just want to try to reproduce problem ...

Dialhot · #**105** 01-10-2006, 04:02 PM

Quote:

Originally Posted by Sagittaire

I just want to try to reproduce problem ...

Do not reduce, do cancel it : stop using mencoder

Sagittaire · #**106** 02-24-2006, 09:16 AM

Quote:

Yes. I'm pissed of discussions about quality of encoder based on metrics

Little experience from MSU for convergence between objective an subjectives test:
http://www.compression.ru/video/code...parison_en.pdf

MOS is average subjective mark
SSIM is HVS tuned metric from video quality experts group
http://www.cns.nyu.edu/~zwang/files/papers/vssim.pdf

Dialhot · #**107** 02-24-2006, 10:52 AM

Very interresting study. Thanks

I will read into it more deeply but I loved the conclusion :

Quote:

• Open source x264 codec of the new H.264 compression standard was found out to be better than long-developed proprietary solutions.
• Subjective comparison is stable and productive method for video systems’ assessment when all testing conditions are precisely adhered to.
• Subjective tests can not always be replaced by objective ones with sufficient precision. SSIM was the best objective metric in our comparison.

May be this time is the last we hear about metrics

Sagittaire · #**108** 02-24-2006, 03:11 PM

Quote:

Originally Posted by Dialhot

Very interresting study. Thanks

I will read into it more deeply but I loved the conclusion :

Quote:

• Open source x264 codec of the new H.264 compression standard was found out to be better than long-developed proprietary solutions.
• Subjective comparison is stable and productive method for video systems’ assessment when all testing conditions are precisely adhered to.
• Subjective tests can not always be replaced by objective ones with sufficient precision. SSIM was the best objective metric in our comparison.

May be this time is the last we hear about metrics

Well MSU will make next MPEG2 test with objective test ...

In fact problem here is not metric but confidence interval:
example: if codec A is better than codec B for 0.1 dB you can't make conclusion but if codec A is better than B for 1 dB you can say codec A is better than codec B with 99% of confidence.

Quote:

Originally Posted by MSU

As you can see, PSNR and VQM provided prediction of similar quality on our test set, and it was quite poor, meanwhile SSIM reached prediction close to the ideal one.

Conclusion : metric are not perfect but you can use it for video quality evaluation and MSU will make it for the next MPEG2 encoder comparison ...

Sagittaire · #**109** 02-24-2006, 03:23 PM

Quote:

Originally Posted by incredible

I think the future would be nice when standarding h264 and aac in that mp4 container.

Here MPEG4 technology demo for the next SAP generation(Certainely H264 HP@L4 compatible)

Superman
Video MPEG4 AVC High Profil 1920*1088*23.976 1600 Kbps
Audio MPEG4 AAC High Efficiency 5.1 48 Khz 160 Kbps

XMenIII
Video MPEG4 AVC High Profil 1920*1088*23.976 1600 Kbps
Audio MPEG4 AAC High Efficiency 2.0 48 Khz 72 Kbps

Perhaps 720p movie on simple 700 Mo CDR if H264 encoder progress always like that ... lol