1080p30 from 1080i60 Conversion

[easy2Bcheesy]

Hi there,

I'm looking at some 1080i60 footage here of a video game that runs at 30fps. In an ideal world, the capture would have locked on at the 'right' field and each 1080i30 frame would be a 1080p30 image - but unfortunately it is not to be.

We have information from one progressive frame spread out over two interlaced frames. I'd like to reconstruct the 1080p30 'original' by reconstituting the image from the 1080i60 footage.

Is this possible? I guess it's just a case of sorting the offset?

[creaothceann]

Got an example clip?

[colours]

DoubleWeave().SelectOdd()

[easy2Bcheesy]

Didn't work, sorry! Here's a clip:

https://www.mediafire.com/?8pscfxtr8zr9ce1

Help much appreciated :)

[feisty2]

simple enough

SeparateFields()
c=last
a=selecteven ().trim (1,c.framecount ()/2-1).assumefps (30.000)+trim (c.framecount ()/2-1,-1).assumefps (30.000)
b=selectodd ().assumefps (30.000)
interleave (a,b).assumefps (30.000)
weave ()

[easy2Bcheesy]

Just wondering if you tried that on the clip? Doesn't seem to help for me? However, I am trying it on an AVI conversion I made of the original .ts so I'm wondering if there's something off with my AVI.

[feisty2]

yeah, I just tried it on your ts sample, it works perfectly

[TheSkiller]

However, if the field shift changes, say if the game duplicates a field or drops one you need TFM.


When I capture off the PS2 where most games run at the frame rate (rather than field rate) of the TV system (25 or 29.97 fps) but are output, of course, in interlaced fashion, TFM is essential because the field shift changes constantly throughout the capture.

Your sample is too short to say it for certain but I wouln't be surprised if the field shift changes from time to time.

TFM(pp=0)

[easy2Bcheesy]

Yeah that was my concern but amazingly it works across the entire segment. Many, many thanks :)

[easy2Bcheesy]

Actually I was wrong - the field ordering does change :/

Where would TFM(pp=0) fit into Feisty2's script?

Another alternative would two exports with slight variations of the original script?

[TheSkiller]

Where would TFM(pp=0) fit into Feisty2's script?It would actually replace feisty's script.

But it's mandatory to set the field order properly.
So, try this:


AssumeTFF()
TFM(pp=0)


Except for the first frame where the matching field is obviously missing this works perfectly with your sample and should work for the rest of the capture as well.


Another alternative would two exports with slight variations of the original script?That would work but it's tedious and unless you visually check the entire capture you never know how many field shifts there are and where.

[easy2Bcheesy]

Your script helps, but there are still several sections that don't work. What a weird vid this one is.

What would the 'inverse' of Feisty2's script be?

[feisty2]

nah, the script at #5 only works on your ts sample, if the field shift pattern of your clip just changes in spurts, I guess TheSkiller's script would be better, and if there's some interlaced mess left behind, you can use "ovr" parameter in "tfm" function to match those frames manually
try this, lower speed but higher precision

tfm (mode=3, slow=2, pp=0)

[TheSkiller]

Your script helps, but there are still several sections that don't work. What a weird vid this one is.Possibly some parts of the game run at a frame rate greater than 30 fps. It's pretty common for menus, maps, inventorys and the like to be rendered at twice the in-game fps, meaning there are no longer pairs of matching fields which make up a progressive frame. These parts would require deinterlacing. TFM can detect those parts and deinterlace them. You can also specify an external deinterlacer for TFM to take the missing lines from since TFM's own interpolation is (imo) not so good.

Yeah, game captures can be nasty to deal with, I know. :D