advanced (2:3:3:2) pulldown removal

[socalver]

Hello,

I know that virtualdub is a great tool for removing 2:3 pulldown from dv avi video, but what will work with the 2:3:3:2 pulldown pattern - other than professional editing software? From what I have been able to read, it seems avisynth might work, but I am not familiar with that program. Is there a plugin for virtualdub, or a program like it, that might work? My version of virtualdub says it only does 2:3 removal. I searched the forums and found https://forum.videohelp.com/topic305663.html?highlight=advanced%20pulldown%20removal this post, which contains a lot of good information, but still need help with the advanced pulldown issue. Any advice would be much appreciated. [/url]

[dphirschler]

If the pattern remains 2:3:3:2 steadily, then AviSynth can easily deal with it. Also look into the Decomb filter for automatic removal, although I prefer to do it manually if possible.

I had a video some time back that I think was 2:3:3:2. It turned out to be animated at 25 FPS and telecined with that pattern to make it broadcast-ready for NTSC. It was strange to me because the animation originated from a Canadian company (Mainstream). Anyways, I wound up using Decomb on it, then I authored it as a PAL DVD.


Darryl

[dphirschler]

Woops. I think my 25 FPS video was 4:2 or something similar.


Darryl

[jagabo]

You can probably do it in VirtualDub with simple decimation. But you have to start the decimation on the right frame and there have to be no pattern breaks.

The point of using 2:3:3:2 pulldown is to have only one interlaced frame out of every 5 frames. The other four are the four unique progressive frames of the 23.976 fps video. When you decimate from 29.97 fps to 23.976 fps in VirtualDub it will throw out one of every 5 frames. The trick is to get it to throw out the interlaced frame, not one of the progressive frames.

Open your file. Go to Video -> Frame Rate. In the Frame Rate Conversion section put a checkmark in Convert To FPS and enter 23.976. Save a few seconds of video starting at frame 0. Check the output file. If it has interlaced frames go back and save starting at frame 1. Check that. If interlaced, start at frame 2... etc. By the time you get to 5, one of the results should have no interlaced frames.

If there are breaks in the pulldown pattern you will start getting interlaced frames again. Even if there are no breaks you will probably eventually get interlaced frames because 23.976 is 100 percent accurate. You might try 23.976024 for a little more accuracy.

[socalver]

If the pattern remains 2:3:3:2 steadily, then AviSynth can easily deal with it. Also look into the Decomb filter for automatic removal, although I prefer to do it manually if possible...

Thank you very much for the advice. I am not familiar with avisynth at all, unfortunately, and all the examples of pulldown removal I have seen through searching are using it for the 2:3 variety. I am sure it is an easy thing to do if you are familiar with that software, which it seems I need to be. Is the 2:3:3:2 pattern just not used very often?

If I am able to do this correctly, will the resulting avi file be 23.976 and progressive when checked by software like media info? When attempting 2:3 removal, it still registered as interlaced. A friend checked the footage in Vegas 9 pro and said the pulldown pattern was 2:3:3:2 - and Vegas movie studio platinum, which I have, won't do pulldown removal.

[socalver]

You can probably do it in VirtualDub with simple decimation. But you have to start the decimation on the right frame and there have to be no pattern breaks...

Thank you very much for this advice. I assume that, for this to work best, the avi segments should be captured with scene detection, as the pattern probably changes with every scene change? And does the audio need any correction with this method?

[jagabo]

You can probably do it in VirtualDub with simple decimation. But you have to start the decimation on the right frame and there have to be no pattern breaks...

Thank you very much for this advice. I assume that, for this to work best, the avi segments should be captured with scene detection, as the pattern probably changes with every scene change? And does the audio need any correction with this method?

Yes, I believe each shot will start a new pattern. Unless the camera automatically stops recording only at the end of a 2:3:3:2 sequence. It's worth a try.

I just ran a little experiment. I took a numbered sequence of 29.97 fps video and decimated to 23.976 fps. VirtualDub threw out every 5th frame. So the original sequence:

0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10

became:

0, 1, 2, 3, 5, 6, 7, 8, 10

The discarded frame was the same for the full minute of my test video. Ie, always frames xxx4, and xxx9.

I believe your 2:3:3:2 video will have the interlaced frame at number 2 and 7 in that sequence. So you need to mark in at frame number 3. Or you can paste 2 frames at the start of the video and mark in at the start.

In AviSynth the command sequence you would use is:

AviSource("filename.avi")
SelectEvery(5,0,1,3,4)

That says, for every 5 frames (number 0 to 4) keep frames 0, 1, 3, and 4. The advantage here is that you don't have to trim away or pad a few frames at the start.