Deinterlacing and MPEG-2

[bodyjarrocks]

Hi, does anyone know what level of deinterlacing is available on this hardware and is it available in this port of xbmc.

According to the Git sources Amlplayer has some support:

Code: Select all

void CAMLPlayer::GetDeinterlaceMethods(Features* deinterlaceMethods)
{
  deinterlaceMethods->push_back(VS_INTERLACEMETHOD_DEINTERLACE);
  return;
}
void CAMLPlayer::GetDeinterlaceModes(Features* deinterlaceModes)
{
  deinterlaceModes->push_back(VS_DEINTERLACEMODE_AUTO);
  return;
}


What does this mean compared to VDPAU modes on Nvidia hardware. Quality de-interlacing methods are critical for watching Australian Rules Football

This also brings me to my second question. Is mpeg2 inside a transport stream hardware accelerated on this combo? In otherwords is DVB at 576i or 1080i going to look as good as on Nvidia h/w or even compared to what my plasma can do itself?

Support of these two is critical for take up of this product in Australia IMHO. Others may disagree and are quite welcome to. AppleTV2 cannot to do this.

Mike (MythTV fan and fan of not wasting electricity redundancy transcending free to air TV to H.264 just to watch it on a low power device.)

[davilla]

libamlplayer will automatically deinterlace such content.

[emveepee]

In the M6 specs Amlogic are advertising ""Direct to panel connection with advanced scaling, de-interlacing and picture
quality enhancement" Are you sure that the interlacing we already have good h/w de-interlacing now?

Martin

[davilla]

This is an M1 not an M6. And yes, it seems to deinterlace fine.

[Hacker-CB]

I'm working on the custom player for the Amlogic M3 platform (ref16).

When I try to play IPTV MPEG2 or H264 streams with 720x576 resolution, picture looks bad. Seems that picture was force de-interlaced by the h/w.

"/sys/class/video/frame_format" during playback is "progressive", but seems that it should be "interlace-top" or "interlace-bottom"...

Is there any way to disable deinterlacing, or some other way to make picture more smoothness?

I'm using the next configuration:

Code: Select all

echo 1080i > /sys/class/display/mode
echo 1 > /sys/class/graphics/fb0/free_scale
echo 1 > /sys/class/ppmgr/ppscale
echo 0 0 1919 1079 >  /sys/class/graphics/fb0/free_scale_axis
echo 0 0 1919 1079 >  /sys/class/video/axis
echo 1920 1080 > /sys/class/ppmgr/disp
echo 0 0 1919 1079 0 >  /sys/class/ppmgr/ppscaler_rect

full /sys/class/video content during playback:

Code: Select all

#for F in `ls /sys/class/video`; do echo $F":"; echo "     "`cat $F`; done   
axis:
     0 0 1919 1079
blackout_policy:
     1
brightness:
     0
contrast:
     0
crop:
     0 0 1920 1080
device_resolution:
     1920x1080
disable_video:
     0
frame_addr:
     0x8bf00000-0x85200000-0x85200000
frame_aspect_ratio:
     0x1c7
frame_canvas_height:
     1088-1088-1088
frame_canvas_width:
     5760-1920-1920
frame_format:
     progressive
frame_height:
     1080
frame_rate:
     Frame rate is 59, and the panel refresh rate is 60, duration is: 2146
frame_width:
     1920
saturation:
     128
screen_mode:
     0:normal
trickmode_duration:
     trickmode frame duration 0
vframe_states:
     vframe_pool_size=4 vframe buf_free_num=0 vframe buf_recycle_num=0 vframe buf_avail_num=3
zoom:
     100

[davilla]

your display is 1080i , why ?


to disable all deinterlacing,

echo 1 > /sys/module/di/parameters/bypass_all

[Hacker-CB]

your display is 1080i , why ?
to disable all deinterlacing,
echo 1 > /sys/module/di/parameters/bypass_all

I have tried 1080i just for test. When I use 1080p picture quality is the same bad.
Your solution "/sys/module/di/parameters/bypass_all" not helps.
May be it not helps because deinterlace module is not enabled in vfm path?

Code: Select all

#cat /sys/class/vfm/map
default_osd { osd(0) amvideo}
default { decoder(1) ppmgr(1) amvideo}
default_ext { vdin(1) amvideo2}

provider list:
   vdin
   decoder.mpeg12
   ppmgr

receiver list:
   ppmgr
   amvideo
   amvideo4osd
   amvideo2
   deinterlace


So, I don't know why "/sys/class/video/frame_format" is "progressive".

I have tried to play the same TS file on the Android:
1. Picture looks more fine, but seems that it has some interlaced vibrations.
2. "/sys/class/video/frame_format" is "interlace-top".
3. No "deinterlace" in the vfm path (like my Linux)
4. My picture from Linux looks like Android picture, when it was paused.

[davilla]

mediainfo on the video file you are trying to play, is it really interlaced content ?

[Hacker-CB]

mediainfo on the video file you are trying to play, is it really interlaced content ?

Yes, media is really interlaced. All media which I has tried was grabbed from the DVB-S Air.

Code: Select all

# ffprobe -show_frames udp://239.195.0.4:1234|grep interlaced_frame
FFprobe version SVN-r92, Copyright (c) 2007-2009 Stefano Sabatini
  libavutil     49.15. 0 / 49.15. 0
  libavcodec    52.20. 0 / 52.20. 1
  libavformat   52.31. 0 / 52.31. 0
  built on Jan 30 2010 11:08:30, gcc: 4.4.3
Input #0, mpegts, from 'udp://239.195.0.4:1234':
  Duration: N/A, start: 62696.465489, bitrate: 15192 kb/s
  Program 10307
    Stream #0.0[0x146]: Video: mpeg2video, yuv420p, 720x576 [PAR 16:15 DAR 4:3], 15000 kb/s, 25 tbr, 90k tbn, 50 tbc
    Stream #0.1[0x19c](rus): Audio: mp2, 48000 Hz, stereo, s16, 192 kb/s
interlaced_frame=1
interlaced_frame=1
interlaced_frame=1
interlaced_frame=1
interlaced_frame=1
interlaced_frame=1
interlaced_frame=1

[Hacker-CB]

I have found the problem. It was appeared after the next Amlogic GIT commit:
commit 416c7dc65d6252d2180d4a47ba903806ad6ec5ac
Author: lei.qian <lei.qian>
Date: Thu Jan 10 14:18:21 2013 +0800
Fix Video bug 65011.

Commit diff:

Code: Select all

diff --git a/drivers/amlogic/amports/vmpeg12.c b/drivers/amlogic/amports/vmpeg12.c
index 98eb7c4..b28bdde 100755
--- a/drivers/amlogic/amports/vmpeg12.c
+++ b/drivers/amlogic/amports/vmpeg12.c
@@ -348,7 +348,7 @@ static irqreturn_t vmpeg12_isr(int irq, void *dev_id)
 
             vf->index = index;
             vf->type = (info & PICINFO_TOP_FIRST) ?
-                       VIDTYPE_INTERLACE_BOTTOM : VIDTYPE_INTERLACE_TOP;
+                       VIDTYPE_INTERLACE_TOP : VIDTYPE_INTERLACE_BOTTOM;
             vf->duration >>= 1;
             vf->duration_pulldown = (info & PICINFO_RPT_FIRST) ?
                                     vf->duration >> 1 : 0;


Video seems better after I was revert this commit
Test interlaced video sample to reproduce this bug: http://update.netsol.su/tmp/samples/rbk-movingline.ts