[FFmpeg-devel] H.264 + PAFF: BBC HD recording shows extreme interlacing artefacts

Jeff Downs heydowns
Tue Nov 6 18:25:36 CET 2007


On Mon, 5 Nov 2007, Reinhard Nissl wrote:
> 
> Your equation matched the frame picture case on ASTRA HD+, but not the
> field picture case, nor the MBAFF case (at least the latter was expected
> behavior).

My analysis is not matching what you are seeing at runtime because the 
premise that the field pair's POC is being determined properly 
(min(top_poc, bottom_poc)) doesn't hold true. See below.


 
> A short excerpt from an above mentioned frame / field / frame
> transition. The message is logged at the position where I've added the
> top_field_first logic. The lines marked with ~~~ indicate, where your
> equation differs from mine.
> 
> picture_structure: 3, field_poc[0]: 164, field_poc[1]: 164, poc: 164
> picture_structure: 3, field_poc[0]: 166, field_poc[1]: 166, poc: 166
> picture_structure: 2, field_poc[0]: 172, field_poc[1]: 173, poc: 173 ~~~

This is a bug for the field pictures. Thanks for looking at this and 
finding it.  Clearly, poc should be 172 here.

Will post patch shortly.

You'll want to re-approach the top_field_first setup with this in mind.

	-Jeff





More information about the ffmpeg-devel mailing list