[MEncoder-users] H.264 compression

Wed May 14 18:23:06 CEST 2008

Em Qua, 2008-05-14 às 10:07 +0100, Mark Himsley escreveu:
> Hi,
> 
> For 6 months or so I've been using the lavcopts in the command below to 
> make rather nice quality H.264 video. Recently, after the update from 
> Ubuntu Gusty to Hardy broke mencoder, I recompiled libx264 and 
> mplayer/mencoder and the files produced are about half the size that 
> they used to be.
> 
> Previously mencoder produced H.264 at about 760 Mbps, now it creates 
> H.264 at about 360 Mbps. To my eyes I feel that the quality has reduced, 
> but I thought that setting vq* to 2 should give me a constant quality.
> 
> It might be that, since I know the files are smaller I think the quality 
> is worse - us humans are odd like that - but I'm wondering if I've 
> overlooked a change in the last few months that would mean my lavcopts 
> less than optimal.
> 
> The full command I am using is (please excuse the line wrapping):
> 
>   mencoder -v \
>    "$i" \
>    -o "$d" \
>    -aspect 1050:576 \
>    -ofps 25.00 \
>    -noskip \
>    -mc 0 \
>    -vfm ffmpeg \
>    -vf kerndeint=18:0:0:1:0,crop=688:568:16:4,scale=480:272,harddup \
>    -ovc lavc \
>    -oac lavc \
>    -lavcopts 
> vglobal=1:vcodec=libx264:vqscale=2:vqmin=2:level=30:mbd=2:trell:v4mv:last_pred=2:dia=-1:vmax_b_frames=0:cmp=3:subcmp=3:autoaspect:aglobal=1:acodec=libfaac:abitrate=128 
> \
>    -of lavf \
>    -lavfopts format=ipod
> 
> 
> And here is the typical output (I pressed ^C after a couple of seconds 
> of encoding)
> 
> 
> MEncoder dev-SVN-r26759-4.2.3 (C) 2000-2008 MPlayer Team
> CPU: Intel(R) Pentium(R) 4 CPU 3.20GHz (Family: 15, Model: 2, Stepping: 9)
> CPUflags: Type: 15 MMX: 1 MMX2: 1 3DNow: 0 3DNow2: 0 SSE: 1 SSE2: 1
> Compiled for x86 CPU with extensions: MMX MMX2 SSE SSE2
> 
> Configuration: --enable-gui --enable-menu
> init_freetype
> get_path('font/font.desc') -> '/home/mdsh/.mplayer/font/font.desc'
> font: can't open file: /home/mdsh/.mplayer/font/font.desc
> font: can't open file: /usr/local/share/mplayer/font/font.desc
> Using MMX (with tiny bit MMX2) Optimized OnScreenDisplay
> [file] File size is 1558312960 bytes
> STREAM: [file] input.mpg
> STREAM: Description: File
> STREAM: Author: Albeu
> STREAM: Comment: based on the code from ??? (probably Arpi)
> success: format: 0  data: 0x0 - 0x5ce1f800
> LAVF_check: MPEG PS format
> Checking for YUV4MPEG2
> ASF_check: not ASF guid!
> Checking for NuppelVideo
> Checking for REAL
> Checking for SMJPEG
> Searching demuxer type for filename input.mpg ext: .mpg
> Checking for Nullsoft Streaming Video
> Checking for MOV
> Checking for VIVO
> header block 1 size: 0
> AVS: avs_check_file - attempting to open file input.mpg
> AVS: File is too big, aborting...
> Checking for PVA
> Checking for MPEG-TS...
> TRIED UP TO POSITION 86857, FOUND 47, packet_size= 0, SEEMS A TS? 0
> Checking for LMLM4 Stream Format
> Invalid packet in LMLM4 stream: ch=0 size=1140851708
> LMLM4 Stream Format not found
> system stream synced at 0xD (13)!
> ==> Found video stream: 0
> ==> Found audio stream: 0
> MPEG-PS file format detected.
> Searching for sequence header... OK!
> VIDEO:  MPEG2  720x576  (aspect 3)  25.000 fps  4350.0 kbps (543.8 kbyte/s)
> [V] filefmt:2  fourcc:0x10000002  size:720x576  fps:25.000  ftime:=0.0400
> ==========================================================================
> Opening audio decoder: [mp3lib] MPEG layer-2, layer-3
> dec_audio: Allocating 4608 + 65536 = 70144 bytes for output buffer.
> mp3lib: using SSE optimized decore!
> MP3lib: init layer2&3 finished, tables done
> MPEG 1.0, Layer II, 48000 Hz 256 kbit Stereo, BPF: 768
> Channels: 2, copyright: No, original: No, CRC: Yes, emphasis: 0
> AUDIO: 48000 Hz, 2 ch, s16le, 256.0 kbit/16.67% (ratio: 32000->192000)
> Selected audio codec: [mp3] afm: mp3lib (mp3lib MPEG layer-2, layer-3)
> ==========================================================================
> [file] File size is 0 bytes
> STREAM: [file] input.mp4
> STREAM: Description: File
> STREAM: Author: Albeu
> STREAM: Comment: based on the code from ??? (probably Arpi)
> ** MUXER_LAVF 
> *****************************************************************
> REMEMBER: MEncoder's libavformat muxing is presently broken and can generate
> INCORRECT files in the presence of B-frames. Moreover, due to bugs MPlayer
> will play these INCORRECT files as if nothing were wrong!
> *******************************************************************************
> OK, exit.
> ALLOCATED STREAM N. 1, type=0
> Opening video filter: [expand osd=1]
> Expand: -1 x -1, -1 ; -1, osd: 1, aspect: 0.000000, round: 1
> Opening video filter: [harddup]
> Opening video filter: [scale w=480 h=272]
> SwScale params: 480 x 272 (-1=no scaling)
> Opening video filter: [crop w=688 h=568 x=16 y=4]
> Crop: 688 x 568, 16 ; 4
> Opening video filter: [kerndeint=18:0:0:1:0]
> ==========================================================================
> Trying to force video codec driver family ffmpeg...
> Opening video decoder: [ffmpeg] FFmpeg's libavcodec codec family
> INFO: libavcodec init OK!
> Selected video codec: [ffmpeg2] vfm: ffmpeg (FFmpeg MPEG-2)
> ==========================================================================
> ALLOCATED STREAM N. 2, type=1
> Building audio filter chain for 48000Hz/2ch/s16le -> 0Hz/0ch/??...
> [libaf] Adding filter dummy
> [dummy] Was reinitialized: 48000Hz/2ch/s16le
> [libaf] Adding filter format
> [format] Changing sample format from little-endian 16-bit signed int to 
> big-endian 8-bit signed int
> [dummy] Was reinitialized: 48000Hz/2ch/s16le
> [format] Changing sample format from little-endian 16-bit signed int to 
> big-endian 8-bit signed int
> Building audio filter chain for 48000Hz/2ch/s16le -> 48000Hz/2ch/s16le...
> [dummy] Was reinitialized: 48000Hz/2ch/s16le
> [format] Changing sample format from little-endian 16-bit signed int to 
> big-endian 8-bit signed int
> [libaf] Removing filter format
> [dummy] Was reinitialized: 48000Hz/2ch/s16le
> Increasing filtered audio buffer size from 0 to 4224
> [ffmpeg] aspect_ratio: 1.777778
> VDec: vo config request - 720 x 576 (preferred colorspace: Planar YV12)
> Trying filter chain: kerndeint crop scale harddup expand lavc
> VDec: using Planar YV12 as output csp (no 0)
> Movie-Aspect is 1.82:1 - prescaling to correct movie aspect.
> VO Config (720x576->1050x576,flags=0,'MPlayer',0x32315659)
> REQ: flags=0x441  req=0x0
> REQ: flags=0x441  req=0x400
> SwScaler: reducing / aligning filtersize 7 -> 8
> SwScaler: reducing / aligning filtersize 7 -> 8
> SwScaler: reducing / aligning filtersize 10 -> 9
> SwScaler: reducing / aligning filtersize 10 -> 9
> [swscaler @ 0x876bb98]BICUBIC scaler, from yuv420p to yuv420p using MMX2
> [swscaler @ 0x876bb98]using 8-tap MMX scaler for horizontal luminance 
> scaling
> [swscaler @ 0x876bb98]using 8-tap MMX scaler for horizontal chrominance 
> scaling
> [swscaler @ 0x876bb98]using n-tap MMX scaler for vertical scaling (YV12 
> like)
> [swscaler @ 0x876bb98]688x568 -> 480x272
> REQ: flags=0x401  req=0x0
> REQ: flags=0x401  req=0x0
> REQ: flags=0x401  req=0x0
> videocodec: libavcodec (480x272 fourcc=34363268 [h264])
> [VE_LAVC] High quality encoding selected (non-realtime)!
> [VE_LAVC] Using constant qscale = 2.000000 (VBR).
> [libx264 @ 0x87833f0]using SAR=1/1
> [libx264 @ 0x87833f0]using cpu capabilities: MMX MMX2 SSE SSE2 Cache64
> *** [kerndeint] Allocating mp_image_t, 720x576x12bpp YUV planar, 622080 
> bytes
> *** [kerndeint] Allocating mp_image_t, 720x576x12bpp YUV planar, 622080 
> bytes5 D/B/S 0/1/0
> *** [crop] Allocating mp_image_t, 720x576x12bpp YUV planar, 622080 bytes
> *** [crop] Allocating mp_image_t, 720x576x12bpp YUV planar, 622080 bytes
> *** [scale] Exporting mp_image_t, 688x568x12bpp YUV planar, 586176 bytes
> *** [harddup] Allocating mp_image_t, 480x272x12bpp YUV planar, 195840 bytes
> *** [expand] Exporting mp_image_t, 480x272x12bpp YUV planar, 195840 bytes
> *** [lavc] Allocating mp_image_t, 480x272x12bpp YUV planar, 195840 bytes
> get_path('subfont.ttf') -> '/home/mdsh/.mplayer/subfont.ttf'
> Unicode font: 255 glyphs.
> get_path('subfont.ttf') -> '/home/mdsh/.mplayer/subfont.ttf'
> Unicode font: 255 glyphs.
> Muxer frame buffer sending 25 frame(s) to the muxer.
> VIDEO CODEC ID: 28
> AUDIO CODEC ID: 15002, TAG: 0
> MUXER_LAVF(audio stream) frame_size: 1024, scale: 1024, sps: 48000, 
> rate: 48000, ctx->block_align = stream->wf->nBlockAlign; 0=1 
> stream->wf->nAvgBytesPerSec:13078
> Writing header...
> *** [kerndeint] Allocating mp_image_t, 720x576x12bpp YUV planar, 622080 
> bytes25 D/B/S 0/1/0
> *** [crop] Allocating mp_image_t, 720x576x12bpp YUV planar, 622080 bytes
> Pos:   3.8s     96f ( 0%)  32fps Trem:  30min  57mb  A-V:0.000 [82:127] 
> A/Vms 2/26 D/B/S 0/1/0
> Flushing video frames.
> Writing index...
> 
> Video stream:   82.345 kbit/s  (10293 B/s)  size: 39114 bytes  3.800 
> secs  96 frames
> 
> Audio stream:  127.423 kbit/s  (15927 B/s)  size: 67959 bytes  4.267 secs
> Uninit audio filters...
> [libaf] Removing filter dummy
> Uninit audio: mp3lib
> Uninit video: ffmpeg
> [libx264 @ 0x87833f0]slice I:1     Avg QP:25.00  size:   432
> [libx264 @ 0x87833f0]slice P:94    Avg QP:28.54  size:   412
> [libx264 @ 0x87833f0]mb I  I16..4: 100.0%  0.0%  0.0%
> [libx264 @ 0x87833f0]mb P  I16..4:  4.3%  0.0%  0.0%  P16..4: 22.0% 
> 0.0%  0.0%  0.0%  0.0%    skip:73.7%
> [libx264 @ 0x87833f0]final ratefactor: -12.14
> [libx264 @ 0x87833f0]SSIM Mean Y:0.9772701
> [libx264 @ 0x87833f0]kb/s:82.3
> 
> 
> Thanks
> 
x264 ratecontrol on FFmpeg (libavcodec) is broken for some time now, for
some weird reason it always use the maximum quantizer. You have to use
x264 directly with mencoder.