dv/avi to x264/mp4 and back to dv/avi

[smok3]

iam looking for a way to backup certain (less important) progresive dv files that would simply take too much disk space in its native form.

questions:
1. Curently iam looking for a fast enough encoding preset for the task, something crf based probably?
2. what would be a good (single step) way to go back to dv/avi? does mencoder includes dv encoder, or some other ideas? (i would like to do everything via cmd/bat)

[Blue_MiSfit]

remember that progressive DV is still stored as interlaced IIRC.

[smok3]

so far i have :

encoding;
x264.exe --crf 18 --bframes 1 --subme 1 --analyse p8x8,b8x8,i4x4,p4x4 --progress --filter "-6,-6"
(i want something faster here)

'decoding';
ffmpeg -i film.mp4 -vcodec dvvideo film.avi

Blue_MiSfit; where would that matter?

[smok3]

this is what i get when comparing original dv with decompressed dv (which went from dv -> x264 -> dv)


Comparing channel(s) RGB

Mean Max Max
Absolute Mean Pos. Neg.
Frame Dev. Dev. Dev. Dev. PSNR (dB)
-----------------------------------------------------
0 0.9743 -0.4510 21 -23 44.3042
1 1.5634 -0.4000 24 -24 40.5992
2 1.2805 -0.4358 19 -21 42.3651
3 1.5170 -0.2722 22 -24 40.9618
4 1.2712 -0.4605 22 -22 42.2621
5 1.4498 -0.3784 21 -21 41.3315
6 1.2222 -0.4441 21 -24 42.6895
7 1.4385 -0.3694 24 -20 41.4082
8 1.2377 -0.4246 20 -21 42.6182
9 1.4509 -0.4076 20 -19 41.3991
10 1.2377 -0.4209 21 -24 42.6394
11 1.4292 -0.3170 23 -22 41.4873
12 1.2350 -0.4362 24 -21 42.6249
13 1.4243 -0.3151 23 -22 41.4967
14 1.3024 -0.4310 20 -23 41.9155
15 1.4454 -0.3360 23 -20 41.3485
16 1.2309 -0.4229 25 -22 42.5094


...
...
...


Total frames processed: 7146

Minimum Average Maximum
Mean Absolute Deviation: 0.9146 1.3070 1.9715
Mean Deviation: -0.5177 -0.3661 -0.1980
PSNR: 37.9991 42.1310 45.0148
Overall PSNR: 42.0593

note: keep in mind that this is RGB comparision.

[foxyshadis]

I don't think it gets much faster than that without --no-chroma-me, --no-cabac, and --nf. But it's practically xvid now, or sub-xvid. Oh well.

What's the size difference?

[smok3]

x264 is like 3-5 times smaller (but i will try some bigger crfs for even smaller file sizes as well), so that could mean instead of backing up 100 gigs, there will be only 20.

well, i have nothing against xvid, but i have all the batches allready done for x264, so...

edit: crf21 gives me:
Total frames processed: 26739

Minimum Average Maximum
Mean Absolute Deviation: 1.0324 1.5683 2.2088
Mean Deviation: -0.5440 -0.3138 -0.0385
PSNR: 37.0949 40.6323 44.0154
Overall PSNR: 40.5588

and subjective test using msu tool:
task type, DSCQS type I
repetitions, 1
average framerate, 1
pause allowed, 1
rewind allowed, 1
one to each, 1
number of tests, 2
number of videos, 3
reference video, C:\luka\dub50_origdv.avi
video 1, video 2, mark of video 1, mark of video 2
C:\luka\dub50_crf21.avi,C:\luka\dub50_origdv.avi,3,5,
C:\luka\dub50_crf18.avi,C:\luka\dub50_origdv.avi,4,5,

Screen resolution, width, 1280, height, 1024,

file, decompressor info,
C:\luka\dub50_origdv.avi,ffdshow Video Codec,
C:\luka\dub50_crf21.avi,ffdshow Video Codec,
C:\luka\dub50_crf18.avi,ffdshow Video Codec,

time of assessment, 13:58 3/7/2006
color blindess test,none

crf21 with this settings and on this material does look quite worse, so ill just keep the crf18 (crf18 is almost identical looking on this old fashioned crt philips tube).

[akupenguin]

x264.exe --crf 18 --bframes 1 --subme 1 --analyse p8x8,b8x8,i4x4,p4x4 --progress --filter "-6,-6"
(i want something faster here)
p4x4 is one of the slow, nearly useless options (except for very low resolution content).

[smok3]

akupenguin, i will try:

x264.exe --crf 18 --bframes 1 --subme 1 --progress --filter "-6,-6"

tommorow.

(that should be faster with almost the same quality no?)

[smok3]

akupenguin, i will try:

x264.exe --crf 18 --bframes 1 --subme 1 --progress --filter "-6,-6"

tommorow.

(that should be faster with almost the same quality no?)

well, encoding speed is the same with this new command line, i guess iam doing something wrong then.... (analize defaults should be ["p8x8,b8x8,i8x8,i4x4"] ?)