[FFmpeg-devel] video corruption using h264 hw decoders

Joakim Plate elupus at ecce.se
Sun Nov 6 01:24:59 CET 2011


On Thu, 3 Nov 2011 16:22:23 +0100, Michael Niedermayer wrote:

> On Thu, Nov 03, 2011 at 12:13:07AM +0100, Joakim Plate wrote:
>> [25 quoted lines suppressed]
> 
> iam not against it if it works but it leaves ambiguity for the left
> and top crop
> 
> 
> [...]

So here's the patch. As you say it does leave an ambuigity for left/top
crop. But that was there before too. Exposing crop using pan_scan could be
an option, but could potentially break compatibility.

pan_scan does have a somewhat different meaning, even if it does fit the
bill rather nicely.

There is one issue with this patch. It doesn't detect changes in crop
values during playback. I'm not sure that matters too much thou seing as
cropping is not fully handled anyway. Also flushing all reference frames on
crop changes seem wrong as well.

fate passes, and i've not seen any odd behaviour in the samples i've tried.

/Joakim

Ps. Any reason those CHROMA444 stuff is there? Why not use croma_x_shift
the same way as croma_y_shift is used?
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-Set-avctx-coded_width-height-to-uncropped-h264-sizes.patch
Type: text/x-patch
Size: 2115 bytes
Desc: Attached file: 0001-Set-avctx-coded_width-height-to-uncropped-h264-sizes.patch
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20111106/47b21c41/attachment.bin>


More information about the ffmpeg-devel mailing list