Paul B Mahol wrote:
On 1/27/16, Andy Furniss <[email protected]> wrote:
Carl Eugen Hoyos wrote:
Andy Furniss <adf.lists <at> gmail.com> writes:
mplayer -ss 11.2 -frames 3 -demuxer lavf -noaspect -vf
ilpack,scale -vo png ~/snooker-short.ts
First convert the input file losslessly to something I-frame only
(like ffv1), then testing will get (much) easier.
Yea was in progress, finished now (used y4m).
So this is the test frame snk-420i.y4m
https://drive.google.com/file/d/0BxP5-S1t9VEENjJpRk5vV0FjSFU/view?usp=sharing
This is what results look like -
https://drive.google.com/file/d/0BxP5-S1t9VEEaEVUTzJQc1dseWc/view?usp=sharing
Isnt your input already 422?
No, that one was a different test I made for zscale so I could compare
with swscale both progressive. Result zscale is better that swscale no
flags but the same if I use flags
-sws_flags +accurate_rnd+full_chroma_int
I guess you could use ffmpeg and scale filter to convert from yuv420
to yuv422p with interl=1 and than use zscale to convert yuv422p to
gbrp and than again scale to convert from gbrp to rgb24 for png:
ffmpeg -i input -vf
scale=interl=1,format=yuv422p,zscale=range=limiter:min=709,format=gbrp
out.png
That's sort of what I did (in more stages) it seems that although going
to 422 interl=1 first is better it still isn't as good as using ilpack.
_______________________________________________
ffmpeg-user mailing list
[email protected]
http://ffmpeg.org/mailman/listinfo/ffmpeg-user
