[FFmpeg-devel] Strange qmax behaviour with mpeg2

[Joseph Artsimovich]

On 12/09/2012 10:41, Michael Niedermayer wrote:
> On Wed, Sep 12, 2012 at 09:57:41AM +0100, Joseph Artsimovich wrote:
>> On 11/09/2012 16:59, Michael Niedermayer wrote:
>>> On Tue, Sep 11, 2012 at 12:02:14PM +0100, Joseph Artsimovich wrote:
>>>> Hi,
>>>>
>>>> When encoding IMX 30 content (I-frame only mpeg2 at fixed bitrate) I
>>>> noticed ffmpeg produces lower quality output than other encoders.
>>>> Digging deeper, I found strange behaviour of qmax parameter with
>>>> mpeg2video encoder that I can only explain away as a bug.
>>>>
>>>> Before explaining further, here is my command line:
>>>> ffmpeg -i input.mov -vtag mx3p -vcodec mpeg2video -r 25 -pix_fmt
>>>> yuv422p -minrate 30000k -maxrate 30000k -b:v 30000k -intra -flags
>>>> +ildct+low_delay -dc 10 -intra_vlc 1 -non_linear_quant 1 -ps 1 -qmin
>>>> 1 -qmax 8 -top 1 -bufsize 1200000 -rc_init_occupancy 1200000
>>>> -rc_buf_aggressivity 0.25 -an out.mov
>>> should be fixed
>> I see this behaviour with the latest ffmpeg from git.
> in that case your input file is needed because the issue is fixed
> with matrixbench_mpeg2.mpg
OK, here you go.

A synthetic .png image to encode [14 KB]: 
http://s17.postimage.org/epzmtwifz/letters.png

Command line1:
ffmpeg -i letters.png -vtag mx3p -vcodec mpeg2video -r 25 -pix_fmt 
yuv422p -minrate 30000k -maxrate 30000k -b:v 30000k -intra -flags 
+ildct+low_delay -dc 10 -intra_vlc 1 -non_linear_quant 1 -ps 1 -qmin 1 
-qmax 1 -lmin QP2LAMBDA -top 1 -bufsize 1200000 -rc_init_occupancy 
1200000 -rc_buf_aggressivity 0.25 -an letters1.mov

Command line2:
ffmpeg -i letters.png -vtag mx3p -vcodec mpeg2video -r 25 -pix_fmt 
yuv422p -minrate 30000k -maxrate 30000k -b:v 30000k -intra -flags 
+ildct+low_delay -dc 10 -intra_vlc 1 -non_linear_quant 1 -ps 1 -qmin 1 
-qmax 12 -lmin QP2LAMBDA -top 1 -bufsize 1200000 -rc_init_occupancy 
1200000 -rc_buf_aggressivity 0.25 -an letters12.mov

The two command lines only differ by qmax and output file name. The 
output files are exactly the same size, which indicates even qmin=qmax=1 
case fits into the specified bitrate. The second file has visible 
compression artifacts, particularly near the bottom. -lmin QP2LAMBDA 
doesn't seem to make any difference.

-- 
Joseph Artsimovich
Senior C++ Applications Developer
MirriAd Ltd