[FFmpeg-user] Optimise XDCAMHD422 Picture Quality - Syntax questions

[Chris von Görstinger]

Hello

I am using this syntax to generate a XDCAMHD422 file:

-c:v mpeg2video -r 25 -pix_fmt yuv422p -s 1920x1080 -aspect 16:9 -minrate
50000k -maxrate 50000k -b:v 50000k -g 12 -flags +ildct+ilme -intra_vlc 1
-dc 10 -non_linear_quant 1 -bf 2 -qmin 1 -qmax 12 -top 1 -bufsize 17825792
-rc_init_occupancy 17825792 -rc_min_vbv_use 1 -rc_max_vbv_use 1
-sc_threshold 1000000000 -lmin "1*QP2LAMBDA" -vtag xd5c -color_range tv
-color_primaries bt709 -color_trc bt709 -c:a pcm_s24le -ar 48000 -f mxf
-timecode 12:34:56:11 -max_muxing_queue_size 1024 -metadata
"creation_time=2015-10-28T15:37:56"

I mentioned by checking frame by frame with an editing tool (after effects)
that there are some frame with very much blockingess.
I tried with qmin/qman different values and surprisingly when using qmin=4
then the quality gets better.

With qmin4 there are still tiny blocks (looks better than qmin=1), and I
now from other transcoders that the frames were these tiny blocks happen
CAN BE better encoded so that no blocks are visible for the whole GOP.
My questions are:
1.) Why do I get a better quality with qmin=4 instead of qmin=1? -> Maybe
is datarate saved with qmin=4 which can be used for other frames were it is
needed more?
2.) How can I improve the encoding a little bit more here in ffmpeg
regarding the picture quality?

br
Christoph
_______________________________________________
ffmpeg-user mailing list
ffmpeg-***@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-user

To unsubscribe, visit link above, or email
ffmpeg-user-requ

[Frédéric Busnel-Joncour]

Post by Chris von Görstinger
[…]
With qmin4 there are still tiny blocks (looks better than qmin=1), and I
now from other transcoders that the frames were these tiny blocks happen
CAN BE better encoded so that no blocks are visible for the whole GOP.
1.) Why do I get a better quality with qmin=4 instead of qmin=1? -> Maybe
is datarate saved with qmin=4 which can be used for other frames were it is
needed more?
2.) How can I improve the encoding a little bit more here in ffmpeg
regarding the picture quality?

Hi there,


These are very interesting questions !
I can't explain the first point, but probably 2-pass encoding can improve encoding the way you want.
If an expert can confirm and give a command line example, that would be perfect.


Best,

Fred

_______________________________________________
ffmpeg-user mailing list
ffmpeg-***@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-user

To unsubscribe, visit link above, or email
ffm

[Marton Balint]

Post by Chris von Görstinger
Hello
-c:v mpeg2video -r 25 -pix_fmt yuv422p -s 1920x1080 -aspect 16:9 -minrate
50000k -maxrate 50000k -b:v 50000k -g 12 -flags +ildct+ilme -intra_vlc 1
-dc 10 -non_linear_quant 1 -bf 2 -qmin 1 -qmax 12 -top 1 -bufsize 17825792
-rc_init_occupancy 17825792 -rc_min_vbv_use 1 -rc_max_vbv_use 1
-sc_threshold 1000000000 -lmin "1*QP2LAMBDA" -vtag xd5c -color_range tv
-color_primaries bt709 -color_trc bt709 -c:a pcm_s24le -ar 48000 -f mxf
-timecode 12:34:56:11 -max_muxing_queue_size 1024 -metadata
"creation_time=2015-10-28T15:37:56"
I mentioned by checking frame by frame with an editing tool (after effects)
that there are some frame with very much blockingess.
I tried with qmin/qman different values and surprisingly when using qmin=4
then the quality gets better.
With qmin4 there are still tiny blocks (looks better than qmin=1), and I
now from other transcoders that the frames were these tiny blocks happen
CAN BE better encoded so that no blocks are visible for the whole GOP.
1.) Why do I get a better quality with qmin=4 instead of qmin=1? -> Maybe
is datarate saved with qmin=4 which can be used for other frames were it is
needed more?

That would be my guess too.

Post by Chris von Görstinger
2.) How can I improve the encoding a little bit more here in ffmpeg
regarding the picture quality?

You can try using -mbd bits, it's going to be significantly slower, but in
my experience it provides better quality.

Regards,
Marton
_______________________________________________
ffmpeg-user mailing list
ffmpeg-***@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-user

To unsubscribe, visit link above, or email
ffmpeg-user-***@ffmpeg.org with subject "u

[Chris von Görstinger]

Hello Marton,

I did use mdb bits and rt. But it didnt get better in my opinion. Maybe I
am too sensible for blocking artefacts.
mdb is my last option to improve the mpeg2 encoding. Right?

Post by Chris von Görstinger

Post by Chris von Görstinger
Hello
-c:v mpeg2video -r 25 -pix_fmt yuv422p -s 1920x1080 -aspect 16:9 -minrate
50000k -maxrate 50000k -b:v 50000k -g 12 -flags +ildct+ilme -intra_vlc 1
-dc 10 -non_linear_quant 1 -bf 2 -qmin 1 -qmax 12 -top 1 -bufsize

17825792

Post by Chris von Görstinger
-rc_init_occupancy 17825792 -rc_min_vbv_use 1 -rc_max_vbv_use 1
-sc_threshold 1000000000 -lmin "1*QP2LAMBDA" -vtag xd5c -color_range tv
-color_primaries bt709 -color_trc bt709 -c:a pcm_s24le -ar 48000 -f mxf
-timecode 12:34:56:11 -max_muxing_queue_size 1024 -metadata
"creation_time=2015-10-28T15:37:56"
I mentioned by checking frame by frame with an editing tool (after

effects)

Post by Chris von Görstinger
that there are some frame with very much blockingess.
I tried with qmin/qman different values and surprisingly when using

qmin=4

Post by Chris von Görstinger
then the quality gets better.
With qmin4 there are still tiny blocks (looks better than qmin=1), and I
now from other transcoders that the frames were these tiny blocks happen
CAN BE better encoded so that no blocks are visible for the whole GOP.
1.) Why do I get a better quality with qmin=4 instead of qmin=1? -> Maybe
is datarate saved with qmin=4 which can be used for other frames were it

is

Post by Chris von Görstinger
needed more?

That would be my guess too.

Post by Chris von Görstinger
2.) How can I improve the encoding a little bit more here in ffmpeg
regarding the picture quality?

You can try using -mbd bits, it's going to be significantly slower, but in
my experience it provides better quality.
Regards,
Marton
_______________________________________________
ffmpeg-user mailing list
http://ffmpeg.org/mailman/listinfo/ffmpeg-user
To unsubscribe, visit link above, or email

--
--------------------------------------------------------------------------------------------------------
Student der Fachhochschule für Telekommunikation und Medien St. Pölten
Student of the university of applied sciences St. Pölten

Christoph Gerstbauer
St. Paul-Gasse 6/4
A-3500 Krems an der Donau
phone: +43 (0) 699 - 11 658 238
email: ***@fh-stpoelten.ac.at
***@gmx.at
***@gmail.com
ICQ: 292233201
_______________________________________________
ffmpeg-user mailing list
ffmpeg-***@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-user

To unsubscribe, visit link above, or email
ffmpeg-user-re