[MPlayer-dev-eng] Paid task
roori bar
roori.bar at gmail.com
Thu Jul 10 03:13:54 CEST 2008
Reimar,
Thanks, it does look similar to MainConcept's decoder, and xvid decoder.
it seems that the fact you don't know how the source looks like, its hard to
estimate what the goal is.
so let me try to clarify that, the current corruption that being shown.. is
usually being produced due to lack of bandwidth (similar to).
my samples do not suffer from these corruptions, due to the fact its being
encoded and playbacked by LAN, which means theres practically no bandwidth
dependency.
so as far as a sample that shows the issue clearly - capture2.bin applys,
all of those 'blocks' / 'redrawing' / 'smudge' , do not exist.
btw, i've ran a script on mencoder and ffmpeg, that tries to use all thier
decoders.. and none produced better results.
roori.
On Wed, Jul 9, 2008 at 10:36 PM, Reimar Döffinger <
Reimar.Doeffinger at stud.uni-karlsruhe.de> wrote:
> On Wed, Jul 09, 2008 at 09:28:28PM +0300, roori bar wrote:
> > first of all, I've noticed that capture.bin is to "easy" sample.. which
> does
> > not clearly shows the issue.
> > I've provided another sample that shows it much better:
> > http://iii.dontexist.net/capture2.bin
>
> Well, those big problems have an obvious fix, I was hoping for a sample
> that shows the problems really clearly even after applying that patch
> (since I refuse to execute some dubios Windows binary blobs I do not
> know how much better it is supposed to look, though I expect it to be
> the same as MainConcept after this change):
>
> Index: libavcodec/h263.c
> ===================================================================
> --- libavcodec/h263.c (revision 14094)
> +++ libavcodec/h263.c (working copy)
> @@ -4745,8 +4745,8 @@
> dc_pred_dir= (s->pred_dir_table[s->mb_x +
> s->mb_y*s->mb_stride]<<n)&32;
> }else{
> level = mpeg4_decode_dc(s, n, &dc_pred_dir);
> - if (level < 0)
> - return -1;
> +// if (level < 0)
> +// return -1;
> }
> block[0] = level;
> i = 0;
>
More information about the MPlayer-dev-eng
mailing list