Mug Funky
23rd February 2011, 03:42
hi!
i'm experimenting with x264 r1900 as an intermediate format for uncompressed captures.
one thing i've noticed is when i encode interlaced (both tff and bff), and decoding the result with ffms2 in avisynth is that the chroma appears to be in opposite field-order with respect to the luma (which is perfect :))
i'm currently having trouble seeing if it's a decoder thing, an encoder thing, or something else.
the input going into x264 (with demuxer set to lavf to avoid the indexing pass) is v210 - that is uncompressed 10-bit 4:2:2.
i've tried both NTSC and PAL captures
considering there's decoding and encoding happening all through the chain, and the source is 4:2:2, i'm at a loss to know where the problem is occurring.
any thoughts?
[edit]
it seems that ffms2 in avisynth is the culprit here. ffdshow (using ffmpeg-mt to decode) worked fine.
[edit]
nope, there's definitely a problem.
tried 2 avisynth scripts, both identical except one delivered yuy2 and the other had a converttoyv12(interlaced=true) at the end.
decoding the result and slapping utoy() on it showed the yv12 one had chroma interlacing, the yuy2 one seems to have progressively downsampled to yv12 before encoding.
am i missing some thing in telling x264 what i'm giving it? i thought maybe the --tff or --bff switches would be sufficient, but i suppose that may not necessarily be.
i'm experimenting with x264 r1900 as an intermediate format for uncompressed captures.
one thing i've noticed is when i encode interlaced (both tff and bff), and decoding the result with ffms2 in avisynth is that the chroma appears to be in opposite field-order with respect to the luma (which is perfect :))
i'm currently having trouble seeing if it's a decoder thing, an encoder thing, or something else.
the input going into x264 (with demuxer set to lavf to avoid the indexing pass) is v210 - that is uncompressed 10-bit 4:2:2.
i've tried both NTSC and PAL captures
considering there's decoding and encoding happening all through the chain, and the source is 4:2:2, i'm at a loss to know where the problem is occurring.
any thoughts?
[edit]
it seems that ffms2 in avisynth is the culprit here. ffdshow (using ffmpeg-mt to decode) worked fine.
[edit]
nope, there's definitely a problem.
tried 2 avisynth scripts, both identical except one delivered yuy2 and the other had a converttoyv12(interlaced=true) at the end.
decoding the result and slapping utoy() on it showed the yv12 one had chroma interlacing, the yuy2 one seems to have progressively downsampled to yv12 before encoding.
am i missing some thing in telling x264 what i'm giving it? i thought maybe the --tff or --bff switches would be sufficient, but i suppose that may not necessarily be.