[x264-devel] constrained intra prediction

[Purvin Pandit]

Hi

Is constrained intra prediction supported in x264? I quick search for the text in version (x264-snapshot-20080824-2245) shows that it is set to '0' in the PPS but I wasnt sure it this is ever used.

Additionally I changed the code to set it to '0' and '1' and encoded a sequence. The resulting reconstructed YUV file has the same md5sum which seems to indicate that contrained intra prediction is not supported.

Can someone confirm this? Also if this is not supported then is there a plan to add support?

Thanks,
-Purvin
_________________________________________________________________
Windows Live?: Keep your life in sync.
http://windowslive.com/howitworks?ocid=TXT_TAGLM_WL_t1_allup_howitworks_012009
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.videolan.org/pipermail/x264-devel/attachments/20090116/a196b577/attachment.htm

[Jason Garrett-Glaser]

Post by Purvin Pandit
Hi
Is constrained intra prediction supported in x264? I quick search for the
text in version (x264-snapshot-20080824-2245) shows that it is set to '0' in
the PPS but I wasnt sure it this is ever used.
Additionally I changed the code to set it to '0' and '1' and encoded a
sequence. The resulting reconstructed YUV file has the same md5sum which
seems to indicate that contrained intra prediction is not supported.
Can someone confirm this? Also if this is not supported then is there a plan
to add support?

No, it is not supported. You are welcome to try to convince us that
there is a useful reason to support it.

Dark Shikari

[Purvin Pandit]

I was looking at x264 as a platform for SVC which requires that the base layer have the constrained intra pred flag set to true.

Thanks,
-Purvin

PS: I am not trying to convince you to implement SVC.

Date: Fri, 16 Jan 2009 18:28:58 -0500> From: darkshikari at gmail.com> To: x264-devel at videolan.org> Subject: Re: [x264-devel] constrained intra prediction> > 2009/1/16 Purvin Pandit <purvinp at hotmail.com>:> > Hi> >> > Is constrained intra prediction supported in x264? I quick search for the> > text in version (x264-snapshot-20080824-2245) shows that it is set to '0' in> > the PPS but I wasnt sure it this is ever used.> >> > Additionally I changed the code to set it to '0' and '1' and encoded a> > sequence. The resulting reconstructed YUV file has the same md5sum which> > seems to indicate that contrained intra prediction is not supported.> >> > Can someone confirm this? Also if this is not supported then is there a plan> > to add support?> > No, it is not supported. You are welcome to try to convince us that> there is a useful reason to support it.> > Dark Shikari> _______________________________________________> x264-devel mailing list> x264-devel at videolan.org> http://mailman.videolan.org/listinfo/x264-devel

_________________________________________________________________
Windows Live?: Keep your life in sync.
http://windowslive.com/howitworks?ocid=TXT_TAGLM_WL_t1_allup_howitworks_012009
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.videolan.org/pipermail/x264-devel/attachments/20090116/67c9a33f/attachment.htm

[Jason Garrett-Glaser]

Post by Purvin Pandit
I was looking at x264 as a platform for SVC which requires that the base
layer have the constrained intra pred flag set to true.

Is there any particular reason why this is the case?

Dark Shikari

[Phil Rutschman]

I'm not following SVC very closely, but here's a reference to that
requirement in an article about it I was reading:

"Additionally, constrained intra prediction
has to be used in the reference layer such that no inter-predicted
samples are employed for intra prediction in the reference layer.
With these mandatory restrictions in SVC, each supported layer
can be decoded with a single motion compensation loop. The
complexity overhead that is required for spatial scalable coding
in SVC in comparison to single-layer coding is smaller than
in the scalable profiles of MPEG-2 Video, H.263, or MPEG-4
Visual."

[Performance Analysis of SVC
Wien, M.; Schwarz, H.; Oelbaum, T.
Circuits and Systems for Video Technology, IEEE Transactions on
Volume 17, Issue 9, Date: Sept. 2007, Pages: 1194 - 1203 ]

Phil Rutschman <philr at modsystems.com>

-----Original Message-----
From: x264-devel-bounces at videolan.org [mailto:x264-devel-
bounces at videolan.org] On Behalf Of Jason Garrett-Glaser
Sent: Friday, January 16, 2009 4:14 PM
To: Mailing list for x264 developers
Subject: Re: [x264-devel] constrained intra prediction

Post by Purvin Pandit
I was looking at x264 as a platform for SVC which requires that the

base

Post by Purvin Pandit
layer have the constrained intra pred flag set to true.

Is there any particular reason why this is the case?
Dark Shikari
_______________________________________________
x264-devel mailing list
x264-devel at videolan.org
http://mailman.videolan.org/listinfo/x264-devel

[Purvin Pandit]

Hi



I would like to encode a sequence with regular intra period (say every 25 frames). I tried using the -i option but that did not seem to work. I used -I option but I think this is generating IDR pictures. Can someone tell me if its possible to generate an AVC bitstream with regualr I period?



Currently I am using the following command line:



x264.exe --scenecut 0 -b 3 --no-b-adapt --b-pyramid -r 1 -i 25 --bitrate 30000 -p 1 --merange 64 -t 2 -v -o test_p1.264 input.yuv 1920x1080



I wasnt sure how to disable scene cut so I use --scenecut 0. Is this conflicting with -i 25?



Thanks,

-Purvin

_________________________________________________________________
Windows Live? Groups: Create an online spot for your favorite groups to meet.
http://windowslive.com/online/groups?ocid=TXT_TAGLM_WL_groups_032009
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x264-devel/attachments/20090309/42d9547a/attachment.htm>

[Jason Garrett-Glaser]

Post by Purvin Pandit
Hi
I would like to encode a sequence with regular intra period (say every 25
frames). I tried using the -i option but that did not seem to work. I used
-I option but I think this is generating IDR pictures. Can someone tell me
if its possible to generate an AVC bitstream with regualr I period?
x264.exe --scenecut 0?-b 3 --no-b-adapt --b-pyramid -r 1 -i 25 --bitrate
30000 -p 1 --merange 64 -t 2 -v -o test_p1.264 input.yuv 1920x1080
I wasnt sure how to disable scene cut so I use --scenecut 0. Is this
conflicting with -i 25?

x264 will only generate non-IDR I-frames in the case that it detects a
scenecut but it has been less than --min-keyint since the previous
keyframe. All non-scenecut I-frames are IDR.

Dark Shikari

[Purvin Pandit]

Thanks for the quick reply. A quick follow up question. With x264 is it possible to encode using fixed GOP size? NO adapation.



I want to encode a sequence with say GOP = 4 (I b B b P), is that possible?



Thanks,

-Purvin

Date: Mon, 9 Mar 2009 11:59:34 -0800
From: darkshikari at gmail.com
To: x264-devel at videolan.org
Subject: Re: [x264-devel] regular intra period

Post by Purvin Pandit
Hi
I would like to encode a sequence with regular intra period (say every 25
frames). I tried using the -i option but that did not seem to work. I used
-I option but I think this is generating IDR pictures. Can someone tell me
if its possible to generate an AVC bitstream with regualr I period?
x264.exe --scenecut 0 -b 3 --no-b-adapt --b-pyramid -r 1 -i 25 --bitrate
30000 -p 1 --merange 64 -t 2 -v -o test_p1.264 input.yuv 1920x1080
I wasnt sure how to disable scene cut so I use --scenecut 0. Is this
conflicting with -i 25?

x264 will only generate non-IDR I-frames in the case that it detects a
scenecut but it has been less than --min-keyint since the previous
keyframe. All non-scenecut I-frames are IDR.
Dark Shikari
_______________________________________________
x264-devel mailing list
x264-devel at videolan.org
http://mailman.videolan.org/listinfo/x264-devel

_________________________________________________________________
Hotmail? is up to 70% faster. Now good news travels really fast.
http://windowslive.com/online/hotmail?ocid=TXT_TAGLM_WL_HM_70faster_032009
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x264-devel/attachments/20090310/8870f3cb/attachment.htm>

[Jason Garrett-Glaser]

Thanks for the quick?reply. A quick follow up question. With x264 is it
possible to encode using fixed GOP size? NO adapation.
I want to encode a?sequence with say GOP =?4 (I?b B b P), is that possible?

--no-scenecut abd --no-b-adapt

Dark Shikari

[Peter.List]

...

Post by Jason Garrett-Glaser
x264 will only generate non-IDR I-frames in the case that it detects a
scenecut but it has been less than --min-keyint since the previous
keyframe. All non-scenecut I-frames are IDR.
Dark Shikari

Hi Jason,
I am wondering, what's the idea behind the fact that scene cuts are NOT
made IDR.

Isn't it true, if I have detected a scene cut I explicitly do not want
any prediction from temporally previous B-frames, because these belong
to a different scene? Isn't it in particular for scene cuts, that I do
want IDR frames?


Regards
Peter

[Jason Garrett-Glaser]

Post by Peter.List
...

Post by Jason Garrett-Glaser
x264 will only generate non-IDR I-frames in the case that it detects a
scenecut but it has been less than --min-keyint since the previous
keyframe. ?All non-scenecut I-frames are IDR.
Dark Shikari

Hi Jason,
I am wondering, what's the idea behind the fact that scene cuts are NOT
made IDR.

That is not what I said. Read my email again, in particular, the
phrase starting with the word "but."

Dark Shikari

[Peter.List]

Post by Jason Garrett-Glaser

Post by Peter.List
...

Post by Jason Garrett-Glaser
x264 will only generate non-IDR I-frames in the case that it detects a
scenecut but it has been less than --min-keyint since the previous
keyframe. ?All non-scenecut I-frames are IDR.
Dark Shikari

Hi Jason,
I am wondering, what's the idea behind the fact that scene cuts are NOT
made IDR.

That is not what I said. Read my email again, in particular, the
phrase starting with the word "but."
Dark Shikari

Ok... I guess the "but" means (logical) AND. I still don't understand though, what's the advantage for a non IDR frame in this case.

Also, for a MPEG2 style GOP-structure with a key frame interval of 24 frames, what would be your preferred number for --min-keyint?

Regards
Peter

[Jason Garrett-Glaser]

Post by Peter.List

Post by Peter.List
...

Post by Jason Garrett-Glaser
x264 will only generate non-IDR I-frames in the case that it detects a
scenecut but it has been less than --min-keyint since the previous
keyframe. ?All non-scenecut I-frames are IDR.
Dark Shikari

Hi Jason,
I am wondering, what's the idea behind the fact that scene cuts are NOT
made IDR.

That is not what I said. ?Read my email again, in particular, the
phrase starting with the word "but."
Dark Shikari

Ok... I guess the "but" means (logical) AND. I still don't understand though, what's the advantage for a non IDR frame in this case.

If the user has specified a minimum keyframe interval that would
prohibit an IDR frame at the current scenecut, the encoder has *NO
CHOICE* but to not use an IDR frame!

Also, see the case of single-frame flashes and such that could
theoretically result in a ton of sequential IDR frames, which would be
bad (and clear the reference list repeatedly).

Post by Peter.List
Also, for a MPEG2 style GOP-structure with a key frame interval of 24 frames, what would be your preferred number for --min-keyint?

Probably 2.

Dark Shikari