Reversing interlaced DV footage

[HighInBC]

I am trying to reverse my footage, it in the DV format, NTSC and interlaced.

The problem is when I reverse the clips the interlacing lines don't work properly and it looks bad.

I have tried the following:

clip.Reverse()

and:

function interlaced_reverse(clip clip1)
{
clip1 = clip1.SeparateFields()
odd = clip1.SelectOdd.Reverse()
evn = clip1.SelectEven.Reverse()

return Interleave(evn,odd).Weave().DoubleWeave.SelectOdd()
}

but both cause different badness. I do not wish to deinterlace as my final destination is television. Please help me.

[HighInBC]

I figured it out! it was simply a matter of deleting one frame from the beggining after seperate feilds, this works(I added '.trim(1,0)'):

function interlaced_reverse(clip clip1)
{
clip1 = clip1.SeparateFields().trim(1,0)
odd = clip1.SelectOdd.Reverse()
evn = clip1.SelectEven.Reverse()

return Interleave(evn,odd).Weave().DoubleWeave.SelectOdd()
}

[neuron2]

Yes, that is one way. Another is to shift the whole picture up by one line (and fill in the new bottom line with a copy). Simon Walters has an Aviynth filter to do that and I have a VirtualDub one. You might prefer it if you want to keep the original fields paired as they are in the source clip. Also, it avoids losing a field at the beginning, affecting audio sync.

[HighInBC]

Thanks, I will remember that if I ever have to work with progressive frames. Once interlacing is properly conceptualized the solutions seem to be very simple.

[neuron2]

Originally posted by HighInBC
Once interlacing is properly conceptualized the solutions seem to be very simple. Shhhhh! That is a secret I've been trying to hide for a long time. :rolleyes:

People do seem to confuse field swapping and field dominance reversal, though. The easy way to think of it is that field swapping is a spatial exchange, and field dominance reversal is a temporal exchange. See here for more details:

http://neuron2.net/reverse/reverse.html

Your method loses a field at the beginning, affecting audio sync.

[HighInBC]

ack! I did not even think about audio synch... a bunch of those can add up. Since I wish to remain in the YUY2 colorspace, I cannot use your vdub filter. I believe this is a fine pure avisynth solution:

function interlaced_reverse(clip clip)
{
a = clip.Crop(0,1,720,479)
b = clip.Crop(0,479,720,1)
return StackVertical(a,b).Reverse()
}

[neuron2]

That will be much slower than Si's native filter. See here:

http://www.geocities.com/siwalters_uk/reversefielddominance.html

[HighInBC]

Done.

[Mug Funky]

this also works:

doubleweave().selectodd()

[stickboy]

Originally posted by neuron2
Your method loses a field at the beginning, affecting audio sync.You also could add a field at the beginning.

Why would audio sync be affected? Does Trim not modify the audio track properly on field-separated clips?

Originally posted by Mug Funky
this also works:

doubleweave().selectodd()DoubleWeave duplicates the last frame, so the last frame won't be quite right.