[FFmpeg-devel] [PATCH] Fix distortion problem in MPEG1/2 video

[Michael Niedermayer]

On Tue, May 14, 2013 at 03:49:08PM +0900, Heesuk Jung wrote:
> Hi,
> 
> I'm sorry for late response and update my feedback for this issue.
> 
> * Background of my suggestion
>  - I believe that many files by pod encoder don't need full parsing option
> and
>    low possibility need full parsing.
>    : As you know and so I concerned performance loss for many normal files.
>  - And I tried to generalize this issue to slove problem ans I realized
> that
>   root cause is wrong frame size information of header.
>    Ans so I suggest that we decide full parsing after checking frame length
> within 7 frames.
> 
> * Full Parsing Enable/Disable speed test
>  1) Test file
>    - I selected high bitrate file to clarify this issue.
>    - Input #0, avi, from
> '/home/hsjung/download/ProblemMedia/AVI/[3]_tp_MPEG2_MP3.avi':
>   Metadata:
>     encoder         : PotEncoder
>   Duration: 00:00:59.23, start: 0.000000, bitrate: 25067 kb/s
>     Stream #0:0: Audio: mp3 (U[0][0][0] / 0x0055), 44100 Hz, stereo, s16p,
> 96 kb/s
>     Stream #0:1: Video: mpeg2video (Main) (MPG2 / 0x3247504D), yuv420p,
> 1280x720 [SAR 1:1 DAR 16:9], 29.97 fps, 29.97 tbr, 29.97 tbn, 59.94 tbc
>  2) Test Environment
>   - Intel(R) Core(TM) i5-3317U CPU @ 1.70GHz
>   - 8GB RAM, 256GB SSD
>  3) Test Method
>   - I evaluated the below excuting time at 5 times.
>   - time ./ffprobe ~/download/ProblemMedia/AVI/\[3\]_tp_MPEG2_MP3.avi
>  4) Evaluate time result
>   - Full Parsing
>             1               2              3              4              5
>            Avg
> real 0m0.051s 0m0.054s 0m0.053s 0m0.056s 0m0.054s 0m0.0536s
> user 0m0.040s 0m0.032s 0m0.036s 0m0.036s 0m0.032s 0m0.0352s
> sys 0m0.008s 0m0.020s 0m0.016s 0m0.016s 0m0.020s 0m0.016s
>   - Header Parsing
>             1               2              3              4              5
>            Avg
> real 0m0.046s 0m0.049s 0m0.047s 0m0.050s 0m0.048s 0m0.048s
> user 0m0.044s 0m0.020s 0m0.024s 0m0.032s 0m0.028s 0m0.0296s
> sys 0m0.004s 0m0.028s 0m0.020s 0m0.016s 0m0.020s 0m0.0176s
>  5) Summary
>   - Test result is about 6ms(Total video duration is 59s) difference
> between Full parsing and header parsing.
> 
> I hope your feedback and thanks

This does not look statistically significant
try yourself: http://www.graphpad.com/quickcalcs/ttest1/

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Democracy is the form of government in which you can choose your dictator
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20130514/4a404327/attachment.asc>