[issue1093] Believed incorrect chroma subsampling on interlaced 4:2:2 to 4:2:0

[Kevin Murray]

New submission from Kevin Murray <kamur...@nds.com>:

I believe there is an issue with the Chroma 
downsampling when going to 4:2:0 with interlaced 
pictures. 

Simplest example is a frame picture containing 
interlaced lines of alternating red and blue (e.g. in 
a ppm). When encoded with mpeg2video or libx264 the 
chroma is smeared and averaged. Forcing subsampling to 
be point or decimation causes one field's chroma values
to be discarded.

A hack to initFilter in swscale.c to generate double
sized filter with alternating zeros, and correct 
offsetting to maintain field chroma polarity start
essentially corrects this (though I'm seeing some
banding of the chroma on each field I would not 
expect).

(FFmpeg version SVN-r18800; libswscale 0.7.1)

----------
messages: 5517
priority: normal
status: new
substatus: new
title: Believed incorrect chroma subsampling on interlaced 4:2:2 to 4:2:0
type: bug

_____________________________________________________
FFmpeg issue tracker <iss...@roundup.ffmpeg.org>
<https://roundup.ffmpeg.org/roundup/ffmpeg/issue1093>
_____________________________________________________