Fields shifting

[Mounir]

I have a problem of fields that are shifting randomly thoughout the video, not sure what's causing this.

Here is a picture (http://img837.imageshack.us/img837/1322/fieldshift.jpg)

and a video sample (164 kb) (http://www.sendspace.com/file/4xfyjb) (ntsc video transfered from vhs)
The shift is occuring on frames 3 and 4

Objectives:
- Keep the video interlaced
- Replace fields in the good order
- No alteration of the OK frames


Thanks for your inputs guys

[johnmeyer]

I was able to fix the problem by adding a AssumeTFF() statement to the beginning of the script.

[Mounir]

assumeTFF() is not solving anything

[poisondeathray]

I have a problem of fields that are shifting randomly thoughout the video, not sure what's causing this.

Here is a picture (http://img837.imageshack.us/img837/1322/fieldshift.jpg)

and a video sample (164 kb) (http://www.sendspace.com/file/4xfyjb) (ntsc video transfered from vhs)
The shift is occuring on frames 3 and 4

Objectives:
- Keep the video interlaced
- Replace fields in the good order
- No alteration of the OK frames



Maybe a dumb question but why can't this be telecine? It looks like music video.

Are you sure it's interlaced ? Doesn't seem so from this short sample. What is the pattern of "random"

Maybe post a longer sample

[johnmeyer]

assumeTFF() is not solving anythingThen try (thanks to Gavino):

AssumeTFF()
SeparateFields().Trim(1,0).Weave()

Depending on how you import the video into AVISynth, you might not need the AssumeTFF() statement (I did need it in my script).

BTW, rather than mess with this small script, you might instead consider re-doing the capture, if you can. I think you used the wrong settings. The capture appears to have been done using some sort of film capture mode in your capture software. This capture software looks like it attempted to do inverse telecine as it captured, but got the fields screwed up. I would capture to a more editable format, rather than compressing to MPEG (this is assuming that you plan to do some editing). Then, I would use the much more capable inverse telecine software available in AVISynth (for instance, IVTC) and let it figure out how to remove pulldown.

And, of course, if you are capture something from VHS, and it is not an off-the-air capture of a film, then I would recommend that you not capture as film-progressive (with inverse telecine), but instead capture the footage as interlaced. I always recommend keeping footage (like TV shows, and VHS video shot with a camera) in its original interlaced format. If you really need to convert it to progressive (for uploading to the Internet, for instance), you can do this later, and do a much better job in AVISynth than what you can do with the software that your capture card runs. My rational for this recommendation is that once you convert an interlaced source to progressive you have now lost, for all time, temporal information that even the best deinterlacer can never replace when the progressive conversion is done.

Therefore, don't do it unless you need to, and then only do it then as the last step in your workflow.

[Mounir]

@johnmeyer
Not quite i have captured it losslessly with huffyuv as i always do. I 'll provide a longer sample if necessary

Edit: here is a longer sample (7mb) (http://www.sendspace.com/file/yyz2af)

[poisondeathray]

Looks like progressive content, telecined. Not surprising for a music video

Mpeg2source()
AssumeTFF()
TFM()
TDecimate()

[Mounir]

Yeah this look better i'd need a pulldown after that any idea?

[poisondeathray]

Yeah this look better i'd need a pulldown after that any idea?


only if you're going back to dvd

you could encode progressive and use soft pulldown

I would clean it up first, and fix the levels (pull down the black level)

[Mounir]

I've got a tougher one (http://www.sendspace.com/file/l7jebu)(64mb sample, from original avi file, huffyuv,vhs) for you Poisondeathray.
In the past i've tried a simple gaussresize but i'm sure i can have something better with your knowledge.
Looking forward to your solution

[manono]

Your capture process is flawed as you should have had a TBC in the chain. One way to improve it is this very slow way:

ConvertToYV12(Interlaced=True)
QTGMC( "Slower", TR0=1, TR1=1, TR2=1)
SelectEven()
TDecimate()

[Mounir]

Thanks for your input, i'll keep it YUY2 though i'm processing the video now hopefully something good come out