Page 1 of 1

More help with Multidecimate

Posted: Mon Jun 17, 2013 9:26 am
by stephen23
I use this lovely filter a lot, mainly to remove duplicate frames from 8mm telecine files prior to stabilising, and have arranged a smooth and efficient workflow.

I am having problems with a file which is not working properly in pass #1: mfile has missing frames at the beginning or the end.

The 8mm scan is done by Cinepost, with a Rank Turbo (truly high-end scanning!), is in DV-AVI PAL, but is interlaced with mismatched fields in VD. Your telecide filter fixes that, and I make and save a DV-AVI file (Cedocida 023) in VD.

I use this saved file in a simple script

avisource("C:\Users\me\Videos\Avisynth\yy12.avi")
converttoyuy2
multidecimate (pass=1)

load into a virgin VD then press Enter or "run video analysis pass", then immediately quit VD.

If I "dub", then mfile has the first few frames after 0 missing, e.g. 1-3 or 1-6. (Then of course the wrong frames are removed)
If I do a video analysis pass instead, the last few frames (e.g. 100 frames!) are missing. (The output stops before the end). VD's "progress" window indicates that these last frames are actually processed.

I thought I'd found a workaround by putting the selection markers at the beginning and end: worked once but now no difference.

Bit mysterious? To do with the Telecide? Be grateful for your help.

Stephen

Re: More help with Multidecimate

Posted: Tue Jun 18, 2013 6:23 am
by admin
Can you provide an unprocessed source sample file that will allow me to duplicate your issue?

Re: More help with Multidecimate

Posted: Tue Jun 18, 2013 1:44 pm
by stephen23
Thanks for swift reply

This is an excerpt - the original had 733 frames.

https://docs.google.com/file/d/0BxpFyO_ ... sp=sharing

Today "run video analysis pass" is working fine, with no loss of frames. Dubbing is variable - always some frames lost immediately after frame #0, between 1 and 10. Have tried trimming the beginning, also saving the file with XVID codec: still reliably drops a few frames.

Re: More help with Multidecimate

Posted: Mon Jul 01, 2013 3:01 pm
by stephen23
Did you reproduce the effect? Here:
Dubbing: always some frames omitted at beginning
Video analysis pass: now OK, no frames lost

Re: More help with Multidecimate

Posted: Mon Jul 01, 2013 5:18 pm
by admin
What settings are you using? If frames at the beginning look like duplicates due to your chosen threshold, then they will be lost.

Re: More help with Multidecimate

Posted: Tue Jul 02, 2013 4:09 am
by stephen23
I haven't run Multidecimate.exe (yet). This is the mfile.txt output from pass #1.
e.g.
0 0.000000
6 3.561256
7 0.692229
8 2.392039
9 3.160094
10 2.075680
11 0.643177
12 2.985938
13 4.780632
14 0.721382
15 2.465293
16 3.994452
17 2.074696
18 0.659269
19 3.883663
...etc
values for frames 1-5 missing.

Re: More help with Multidecimate

Posted: Tue Jul 02, 2013 9:17 pm
by admin
After you open the script in VirtualDub, just do File/Preview input...
Wait for it to play all the way through, then close VirtualDub.

When I do that with your sample I get this:

0 0.000000
1 32.687029
2 11.233306
3 0.665847
4 11.643914
5 2.946577
6 3.561256
7 0.692229
8 2.392039
9 3.160094
10 2.075680
11 0.643177
12 2.985938
13 4.780632
14 0.721382
15 2.465293
16 3.994452
17 2.074696
18 0.659269
19 3.883663
20 2.850370
...

Nothing missing for me.

Re: More help with Multidecimate

Posted: Wed Jul 03, 2013 5:35 am
by stephen23
Thanks for looking.
When I do that it consistently misses frames. And only with that particular avi file or subsections thereof.
It must be something to do with my PC - W7x64 Home Premium, nothing else running. Any idea what sort of thing would do it?

Fortunately "Video analysis pass" now seems to work reliably.

best wishes
Stephen

Re: More help with Multidecimate

Posted: Wed Jul 03, 2013 6:56 am
by admin
That's strange. Can you tell me which version of VirtualDub you are using?

Does it happen with the sample you gave me?

Re: More help with Multidecimate

Posted: Thu Jul 04, 2013 9:13 am
by stephen23
1.9.11, almost certainly X86.

With that sample (and its parent) and no other that I have found. (And I use this filter a lot)