Reversing field dominance

[erratic]

Maybe this question belongs in the NLE forum, but I think I may get a better answer here.

I know you can reverse the field order by shifting the video 1 line up or down, or (probably better?) with these Avisynth commands:

avisource("tff-clip.avi").assumetff.doubleweave.selectodd # for tff video
avisource("bff-clip.avi").doubleweave.selectodd # for bff video

The sharpness won't be affected. However, I can also change the field order with Vegas, and while this works, Vegas seems to do have a strange way of doing it. Click here (http://users.telenet.be/erratic/vegas/field.htm) to see what Vegas did.

Can anyone (maybe an avisynth expert) tell me what Vegas has done? :confused:

[Arachnotron]

Maybe a stupid idea: could it be this was an error made by the zdf? Reversing the field order in the studio *before* letterboxing? So the letterbox is TFF but the video is BFF? In that case anything fixing the video would break the letterbox edge.

Of course, if using avisynth does not give you problems the above is nonsense.

[erratic]

It works fine with Avisynth, so it's not a ZDF error. After changing the field order from tff to bff with Vegas, the field order has effectively changed, so Vegas works but the result is pretty bad. That's why I wonder why Vegas doesn't simply shift the video 1 line up or down, like CCE for example. Vegas appears to use a weird method. I'd rather change the field order with Avisynth before importing the video in Vegas. But fortunately I don't really need to change the field order, it was just an experiment.

[erratic]

It's no longer about Vegas this time, but the question does belong in this thread. Is one of the two methods for reversing field dominance really better than the other?

1. Line shift
2. DoubleWeave().SelectOdd()

I know the difference between these two possibilities and I've read Donald Graft's comments there (http://neuron2.net/reverse/reverse.html). But I don't think it would really make a difference which method you use. The motion of the video on a TV wouldn't be affected either way. Or am I missing something?

I did notice that DoubleWeave().SelectOdd() has an effect on scene changes, though. When the original video cuts from one scene to the next without a transition, you now end up with a 1-frame transition, containing the last field of the first scene and the first field of the second scene. I think this could have a slightly negative effect on frame based mpeg(-1/2/4) encoding.

[scharfis_brain]

vegas video turns the fieldorder this way:

assumetff() # or bff()
bob()
separatefields()
selectevery(4,1,2) # for BFF output or (4,0,3) for TFF output

this means:
- image will blur unnessesary
- but the video doesn't get shifted by 1/50 sec (1/60 sec for NTSC users)

doubleweave().selectodd()
won't introduce blur, but shifts the video by one field in temporal domain.

a one line shift is compared to doubleweave.selectodd, a temporal shift, too, exept, that it shifts the image, what isn't good with YV12 colorspace and macroblock optimisations...

[erratic]

That's indeed what Vegas does, scharfis_brain! But I had to change selectevery(4,1,2) to (4,0,3) to go from tff to bff. The video did indeed become blurrier and the head switching noise line showed up twice. You've impressed me again. :)