On Sun, 4 Jun 2006, E.Chalaron wrote:
> I have changed the y4mscaler to the version 9.0, the problem remains...
> I'll try another size of frame, these values were nice since they did cut off
> the usual crappy border form 8 and super 8.. but hey, I'll survive !
Some of the changes Matt made in y4mscaler 9.0 were specifically aimed
at the problem you were having with larger frame sizes and complex
scaling factors.
I've been unable to reproduce the problem.
> > (find . -type f | sort | xargs cat) | \
I recall you mentioned at one time that the images produced by
the camera were 1280x960.
> > yuyvtoy4m -w 1116 -h 832 -a 1:1 -i p -r 17:1 | \
If the files are 1280x960 then the 'cat' command should produce a
"raw" (headerless) stream YUYV stream
So, I simulated frames from the camera with 'y4mblack':
y4mblack -w 1280 -h 960 -a 1:1 -i p -x 422
and fed that into y4mscaler thus:
y4mblack -w 1280 -h 960 -a 1:1 -i p -x 422 | \
y4mscaler -I active=1116x832+82+64 -O sar=PAL -O size=704x576
I think something like this (to paraphrase your script) would
be close to what you want:
find . -type f | sort | xargs cat) |
yuyvtoy4m -w 1280 -h 960 -a 1:1 -i p -r 17:1 | \
y4mscaler -I active=1116x832+82+64 -O chromass=444 -O sar=PAL -O size=704x576
| \
y4mstabilizer -a 0.7 | \
y4mscaler -O chromass=422 | \
yuvdenoise | \
y4munsharp | \
yuvflip -H | \
yuvcorrect -v 0 -Y LUMINANCE_"$lum"_16_235_16_235 | \
y4mtoqt -o "$reel_name"
NOTE: Do the y4munsharp and yuvdenoise ON THE SMALLER 4:2:2 frames.
It doesn't make sense to denoise the large 4:4:4 frames!
It also makes sense to denoise, unsharp, yuvcorrect the DOWN
scaled frames. Why correct 2 or 3 times as many pixels as
you are going to use? ;)
There might be some work left to do on that sequence of commands
but I think that'll be a LOT faster than the order you originally
were using.
Cheers,
Steven Schultz
_______________________________________________
Mjpeg-users mailing list
Mjpeg-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mjpeg-users
