[FFmpeg-devel] Parallelized h264 proof-of-concept

[Andreas Öman]

Hi

Mathieu Monnier wrote:
>> o deblocking doesn't work correctly. 
> 
> There's a flag (per picture?) indicating whether the deblocking must be 
> done accross slice boudaries or not.
> 

Yep, h264.c sets "h->deblocking_filter = 2" in this case.

I have, however, discovered that filter_mb_fast() does not seem to
comply with this deblocking mode. I assume filter_mb_fast() is supposed
to be bit-exact to filter_mb(), no?

Did some tests with md5 sums and replaced the filter_mb_fast()
with just filter_mb()

File a has deblocking type 0 and 1 (trailer from apple.com)
File b has deblocking type 2 (Discovery HD)
File c has deblocking type 2 (Swedish television HD)

file a: 62f4b715fe34e890b31f09d994320ee2 with filter_mb_fast
file b: 59814559192176465f21fa524cf88604 with filter_mb_fast
file c: 3ea5a957eaa71cfd8d22730af96496e1 with filter_mb_fast

file a: 62f4b715fe34e890b31f09d994320ee2 with filter_mb
file b: 64f5b254ead02014b8d0121b38434ce2 with filter_mb
file c: 7326302eac34979fe9df7f2205790495 with filter_mb

The attached patch remedies this.

file a: 62f4b715fe34e890b31f09d994320ee2 with patch
file b: 64f5b254ead02014b8d0121b38434ce2 with patch
file c: 7326302eac34979fe9df7f2205790495 with patch

-------------- next part --------------
A non-text attachment was scrubbed...
Name: h264.diff
Type: text/x-patch
Size: 739 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20070602/448eab60/attachment.bin>